EBRI_logo_base_black
EBRI

This study examines whether and how HSA balances impact use of and spending on health care services. Most account holders will build up an HSA balance (to the degree that contributions are made) because most people are healthy and do not use a lot of health care services in any given year. Over time, growing HSA balances may mitigate the impact of the deductible. We find that as individuals build up balances in HSAs, they use more health care services than they otherwise would. In essence, HSA balances may blunt the cost-reducing effect of high-deductible health plans over time. The implication is that employers may want to consider the impact of the size of the plan’s deductible. Another approach is to explore education around the long-term benefits of saving the HSA balance for health care expenses in retirement.

The health insurance eligibility and claims data for this study, which come from an employer with approximately 120,000 employees, were matched to HSA balance, contribution, distribution, and investment data from the EBRI HSA Database. A subset of about 6,500 policyholders was used in the study. Data from 2014–2016 were analyzed.

Key findings:

  • In general, we find that larger beginning-of-year HSA balances result in increased use of health care services. Use of the following services increased: emergency department visits, primary care physicians (PCP) and specialist visits, chiropractic services, physical therapy, psychotherapy, blood tests, X-ray, CT scan, MRI, and musculoskeletal surgery.
    • Individuals with $3,000 or more in their HSA at the beginning of the year visited PCPs and specialists on average one-half visit more per year than individuals with HSA balances under $3,000.
    • Spending was $19.52 higher for primary care office visits and $28.06 higher for specialist visits. Overall spending was $496.40 higher for every $1,000 in HSA balances, though the finding was not statistically significant. Spending associated with other outpatient services was $550.96 higher among individuals with an HSA balance of at least $3,000 and was statistically significant.
  • Balances in HSAs more than doubled between 2014 and 2016. Among enrollees with employee-only coverage, balances increased from $1,131 at the beginning of 2014 to $2,861 at the beginning of 2016. Similarly, balances more than doubled among enrollees with family coverage. The average balance was $3,312 at the beginning of 2016.
  • The percentage of accounts with a zero balance at the beginning of the year fell to nearly nothing after two years of contributions. Only 2 percent of HSAs started 2016 with a zero balance. At the beginning of 2014, 20 percent of those with employee-only coverage and 17 percent of those with family coverage had a zero balance. The percentage of accounts with balances between $1,500 and $2,999, and $3,000 or more nearly tripled between the beginning of 2014 and 2016 among those with employee-only coverage. Among those with family coverage, the percentage of accounts with balances between $1,500 and $2,999 more than doubled, while the percentage of accounts with balances of $3,000 or more quadrupled between the beginning of 2014 and 2016.

Figure 13 Figure 12 Figur Figur Figur Figur Figur e 10 e 1 e 7 e 8 e 5 Linear Fixed Effects Model Estimates of HSA Balances on Use of Services Use of Select Outpatient Services, 2014 –2016 Ann Pu er al ce Ind D n HS t is ag ivid A tribu -e Eligib of HS ual tion P Con le e As r of HS He ce tribu With n alth tag A tion Z e Balanc Plan e of HS rs, o Enr b Balanc y T e As a oll ype t In Begin me v e e of Cov a n st Begin tt, en d 2007 in , e g rof Y ag n –in 2018 eg e , ar 2014 of Y , e-ar 2016 , 3 st e Va There Fi R R Hxpla a g e a e v t ur riab sis fe ila ul e t na ic nd, ha r t6, An a s t le e ion o ll n s b nce of D y s u A al N is m si eR te g e r f r uibu n a o nif lia s me b b s ier tus M., c ul ion o lot a on t f ts. Outp tne N of re D f HS ee s riff la atient e rsa se te io jC A r a e Sood ns he B rnc c a hip h condu c la eks s in c , n s c Rola b ee a toe t w nd c B te ff ee e ic d Mc g n HSA ie inn on D nt e ing H v eSA it st btof a , im sla a a a Y nd ns te c ea e w s b M. rs a e , ll P e Sus .nd olic t w The e us a ye n M hold n mod e num of ae rhe q b rs Wi uis e ea rls lt . of h c on t20 h E Ha 11 t SA he rm ea s p . serv ful loy is "T l d b he eic e ea ing e -tE Only s in a aff se t erc ta tC c a b s of k nd ov out ed et C rhe a ta onsu w nnu g o e he - , tahir m ally ltehie d ,rs of - D a rnd ireocur te d Paul Fronstin is director of the Health Research and Education Program at the Employee Benefit Research Institute Do Accumulating HSA Balances Affect Use of Health Care 2014 2015 2016 Policyhold b by T y T er ype ype s With of Cov of Cov Family e er rag agCov e e, , 2014 2014 erag – – e2016 2016 , 2014 –2016 Model 1 Model 2 Services 2014 2015 2016 millions m su infod b or sa e m lm s. ap t ion o H le 2014 e m altight n tr h P –2e la 01 b nds ns e6 e ................................ on v in b ide Ea p nc lis ae od nc of e es, ind s of expe H c iv e tidua e ad lt h he ................................ l a Ca and lt re h se ." e m For rp vic loy um es e rfo us cront e H der ra ibut iv................................ lt ing h Ec ions pono rior , d m is HtS ir cA ibut s a con nd ions tr Pib olic , ut a ................................ nd ions y 14 inv . (e 2) st: m 1e -27. nts is also ........... being 11 (EBRI). M. Christopher Roebuck is president and CEO of RxEconomics, LLC. This Issue Brief was written with assistance Brot-Goldberg, Zarek C., Amitabh Chandra, Benjamin R. Handel, and Jonathan T. Kolstad. 2015. "What Does a A Ws not e exa e 1.0 m d 0ine abov d e a, num our b ee st ri m of ad te es m pende ay nt be vaaff ria ecbtle ed s t bo y b the ett e erndo unde ger ne stit ay nd of the HSA role ba tla hnc at eH . SFA or b e axa lanc me p le h, as on the aus nte ic ip of ahe tion altof h Services and Spending? 0.95 $3 30,25 16 000 % % HSA Balance HSA Balance 20 20 2014 14 14 20 20 2015 15 15 20 20 2016 16 16 tracked (Fronstin, 2018). The research has found that HSA balances tend to build up over time. Furthermore, the from the D Ins edtuc itu tible te’s r D eo? sea The rch Iam nd paecd t itof orC iaost l st-a Sh ffs. aring Anyon vieH wes e altxpress h Caree P dr ic in t es, Qua his rep nt or ittie as, a re tnd hose Spe of nding the a D ut yho narm s ia cs." nd sh NBould ER c us ar ee of serv heia clt eh c s. Ia npa re ser tient vic , e ou s tm pa at yie induc nt, and e pp er op esc ler ipti to c on ont drribut ug serv e to ic the es w ir a ecrc eoun exa tm in a ined dv.a Inc npa e tof ient re cserv eiving ices inc such se lude rv di cte he s. To Figure 7, Distribution of HSA Balance at Beginning of Year, Policyholders With Family Coverage, 2014–2016 ............. 12 ($thousands) ( = $ 3 , 0 0 0) H W aev ila alsnd, o 50e % m Am ploy elia e d M., a nov Neeerl e aj cSood onometr , Rola ic nd tecMc hniD qe ue vit t tha , atnd allo M. wSus s for a n M thea inc rquis lusi . on 2011 of b fix . " eH dow eff e Dco ts w Consu hilem ae lsro -D air de dcrteess d ing He a the lth 0.90 By Paul Fronstin, Ph.D., Employee Benefit Research Institute longer someone has had an HSA, the higher their personal contributions, the more likely they are to invest, and the not be ascribed to the officers, trustees, or other sponsors of EBRI, Employee Benefit Research Institute-Education and Working Paper No. 21632 (National Bureau of Economic Research). Accessed May 20, 2019. 26.1 0.85 num explore bers of this inpa issue tie , we nt hospi ran at anum l adm be is rsi of ons al ta end rnainpa tive tspe ient c if diFi a ca y gs. t uions r e Out 3. p In o atiene nt e se xe rv rc icis ee s inc , welud ree -e dst the im anum ted b ae ll m rs of ode els m e on rge th nc e y May 23, 2 140 %19 • No. 482 14% Dependent Variable 13% d irection o Pla f ns ca u Asa ffe lic ty t Vu (Wlne illia ra m bs, A le Plli op son ula , ta ion nd s? Mor " For al-um Be nit foo r ,H 2018) ealth Ec . Bonomi riefly, c this s a nd rela Ptolic ivey ly 14 ne w (2) m : od 1-12. elin g approach includes $2,479 a mnd ore M. like C ly hris theyt a orpher Ro e to take a e buck, distributPh.D. ion. , RxEconomics, LLC 44% 20% Research Fund http://w (E w Bw RI 0.79 .nb -ERF) er.or , g or /p ta he pe irr s/ stw aff 21 s. N 632.p eithe dfr EBRI nor EBRI-ERF lobbies or takes positions on specific policy Figure 25 8, 45%Annual I 43% ndividual Contributions, by Type of Coverage, 2014-2016............................................................. 13 d su eb psa ar $2 t m m ,5p 00 ele n tof vis indiv its, p idua rimls ar w y it ca h a rR ee a C la nd ha tion rspe lso sh n C ciia ps li om st Be o offi trw bidit e ce e n y v is H Iind StA s,e B c x (C a hir laop n Cce Ir)a of a cn tor d ze H v ro. e isa i tlThi s, p th s e Shy erlisi vm ic ce ina al t s U the esd e r a those py vi si w ts, a ith a nd ny pof syc the hot che hrronic apy 0.80 23.6 person-level fixed effects to control for time-invariant unobserved confounders (as we did in our main analysis), but 20% Inpatient hospital admissions 0.001 $2,310 0.01 proposals. EBRI invites comment on this research. v con isitd s. it ion Othe s trha ou t tcpoam tie pnt ris e out the co C mC eIs a —n a ab lyout zed 36 w ep re er c count ent of s of the la t bot aa nd l sa ra m dp iolog le. Sinc y se er v the icese s — p esu op cle h a as e re m lec or tre o c lik ared ly iog tha ran t ms hose Leive, Adam. 2018. Health Insurance Design Meets Tax Incentives for Saving: Consumer Responses to Complex 12% 40% 38% There a Fils go urus e is e 9, s l aA a q nnu g ue ge st adl D ion a endo istr s t ibut ge o nous w ion he s, b trhe eg yrr Ty e Hss SA por es b s, w of lun Chic ov t te h he he ra gim e lps , p20 aide c1 t 4 nt of –ify 2016 the w high he ................................ the dre rdeuc vetrible se .c a Cus ons alium ty ................................ is er s m at pa la yy v . ie Th wis t he sol a vc ec ............ s a ount n as 13 Inpatient hospital days -0.004 0.05 Buchmueller, Thomas C. 2009. "Consumer-Oriented Health Care Reform St$2,096 rategies: A Review of the Evidence on 0.70 17% ( w Eit Khout G), e c chr hooni carcd c iog ond raim tions s, st troe ss ant tie cst ipa s, X te -trhe ayis, c r fut om ure p u he tea riz lth e dc a tom re c og ost rasp , hy the (y C T) ar esc m aost ns, p and rone m a to gne prt eic -fun resona d the nc ir eH im SAa . g W ing e Contracts. University of Virginia. Accessed May 20, 2019. http://papers.nber.org/conf_papers/f113040.pdf e In Do ea strim m tra Ac a o rtk ion e duc ed for c ff tiion u ccie um nc rr eyu nt p lat rspe oble nding m in in d g , e yspe HS nam cia icA lly p aane Ba nyl d ea m la tp a loy m n od ec r ece ls ont s by r ibut Aff incor ion pe s t orc a o ting t the Us st HrSuc Ae , tu a of rnad l e the q Hea ua 19.3 retfor ion mod e m lt ayh e lin fe Ca e gl th me ar tthod e the s. re A is Suggested Citation: Paul Fronstin and M. Christopher Roebuck. “Do Accumulating HSA Balances Affect Use of Health 20 Manag Ee me d C rom gen pceyt id tion a eparnd tme Cn onsu t vism its er-Directed Health I 0ns .01 urance." * Milbank Qua 0r.t0 e5 rly 87 (4** ): 820-841. $2,000 35% ( found MRI). no Wem aaltso e eriaxa l dm iffine ere dnc if a es b nd ehow twee o n th vere a ll spe result nding s of t, he and se m spe od nding els and fort he the rse esu spe lts cfifi rocm serv thei cful esl sa , wa m s a ple ffe re ctgerd e ss by ion Hs. SA Figure 10, Percentage of HSAs Invested, by Type of Coverage, 2014–2016 ............................................................. 14 10% 10% 0.60 no C rea-re ne est Se e im dr a v tto ic ion o e bs e acf o nd ost uSp rc ons aen nding ac ly ious ses ?” . us W EB ing eRI ha tIhis v ss eue found te c Bhni riefq only , ue no. he tw 48 lpe o 2st d ( E udie us m p ts t o loy fur ha eett he B ee xrne a a m ss fit ine e Re ss t he sea whe im rc the h I parc ns tor t th i tnot a ut t eH , H S Ma S AA s ha y b 23 ala v, e nc 20 on e19 w us ) a.e s a of cause 17.2 Both enrol doi:10.1 lment O u in 11 tpH 1/ aSA ti j.146 e-n et ligi p 8- h b 00 y le s09 ihe ci.2 aa n 00 lt 's h p 9.00 ofla fins c58 e v a 0.x. ind sits the number of 0.H 10 SAs have *** grown sign0 ifi .c 5a 0ntly sinc*** e their first becoming Se rv 15%ices and Spen9% ding? 30% Quan, H., P. Halfon Sundararajan, A. Fong, B. Burnand, J. C. Luthi, L. D. Saunders, C. A. Beck, T. E. Feasby, and W. A. balance. For policyholders with family coverage, the use of health care services variables were aggregated across all 28% 15.7 30% or serv eff ice ec s a t (nd or b spe oth) nding of he . B aot lth h st serv udie ices e s us xa em aine nd dc os the ts. im pact of contributions to HSAs on use of health care services and available in 2004. In 2018, enrollment in HSA-eligible health plans ranged from 23 million to 36.8 million policyholders Figure 11, Use Pof rima Sele ryc c t a H re ea p lth h C ysia cr ie a nSe virsviits ces, 2014–2016 ................................ 0.05 *** ................................ 0.21 ***...................... 14 As mentioned above, we also employed a novel economet 27% ric technique that allows for the inclusion of fixed effects 0.5G 0hali. 2005. "Coding Algorithms for Defining Comorbidities in ICD-9-CM and ICD-10 Administrative Data." individuals covered by the policyholder’s health plan. We did not include preventive services as they are covered in full, By Pa 15ul Fronstin, Ph.D., Employee Benefit Research Institute Copyright Information: This report is copyrighted by the Employee Benefit Research Institute (EBRI). You may copy, Bundo $1r,f, M 58% 00 . Kate. 2012. "Consumer-Directed Health Plans: Do They Deliver?" Research Synthesis Report No. 24 (Robert spending. Leive (2018) found that switching to HSA-eligible health plans did not lower costs, as employees offset the and their dependents (Fronstin, 2019). According to one survey, the number of people enrolled in an HSA-eligible Specialist visits 0.06 *** 0.29 *** while also addressing the direction of causality. This modeling approach adds lagged endogenous regressors to our 25% 7% Medical Care 42 (11): 1130-9. and use of those services should not be expected to be affected by HSA balances. a prnd int, M. or C dow hris nloa to dpher Ro this repore t buck, solely for Ph.D. person , R al a xEco nd nonc nomics omm,e LLC rcial use , provided that all hard copies retain any and Figure 12, Use of Select Outpatient Services, 2014–2016 11.9 ........................................................................................ 15 Wood Johnson Foundation). Accessed May 20, 2019. higher deductibles with HSA contributions. Similarly, Haviland, et al. (2011a) examined variation in employer 11.4 health p 0.4 la 0n increased from 4.9 million to 26.1 million between 2006 and 2018 (Figure 1). And, the number of HSAs model, which he Ch lps iroide prant cti 6% ify c whether reverse causality is at play 0.. 0 I4 n genera l, the results fr 0.2om 9 this e** xercise largely rule 10% $1,064 a Fin ll cop dings yr 6% ight and other applicable notices contained therein, and you may ci1te or quote small portions of 0.33 the report 20ht % tp://www.rwjf.org/content/dam/farm/reports/reports/2012/rwjf402405 contributions to HSAs (and HRAs) a $1,000 nd found that higher employer contributions tempered the effect of the high reached 25.1 million at the end of 2018, containing $53.8 billion (Figure 2). Williams, Richard, Paul D. Allison, and Enrique Moral-Benito. 2018. "Linear Dynamic Panel-Data 0.30 Estimation Using The out rk e 10 e vy e rinde se cp ae u Pnde sa hylistnt iy c.a v I l n fa a th ria eb c ra t le ,p in m y was t any he c pa olic ses y, hold oure or r’s H iginSaA l r beasu lalt nc 0s (posi .e 1 0at the tiv eb e r ** e glinn ation ing shof ips) t0 he .w 4 3 y ee re a rm . W orh e ile pr onounc we cond ed uc us ted in g Figure 13, Linear Fixed Effects Model Estimates of HSA Balances on Use of Servic 17% es ................................................ 16 $1,000 $903 8.6 16% 8.4 0.27 0.30 provided that you do so verbatim and with proper citation. Any use beyond the scope of the foregoing requires EBRI’s deductible on use of health care services. These two studies examined how contributions affect use of health care Trends in HSAs Maximum Likelihood and Structural Equaltion Modeling." The Stata Journal 18 (2): 293-326. a this na ly eses cono us ming etriH c SA ap p bra ola ac nh. ce as a continuous measure, we also examined potential non-linearities in the functional form 15% Psychotherapy 0.05 * 13% -0.10 A T A G L A N C E Buntin, Melinda Beeuwkes , Amelia M. Haviland, Roland McDevitt, and Neeraj Sood. 2011. "Healthcare Spending and prior express 4% permission. For permission 6.2 s, please contact EBRI at permissions@ebri.org. Employers have adopted HSA-eligible health plans as a way to manage the cost of providing health benefits to workers serv Figur icee s. 14 They d , Linear id Finot xed e E xa ffm ecit ne s Mod howe l bE ast lainc me as m tes of ighH t SA aff eBca t la us ne ce of s on hea Sp lth c ending are serv ................................ ices. In addition to t ......................... hese studies, 16 of the r 0.2 ela 0tionship by instead using a dichotomous measure indicating 0.17 whether or not the HSA balance was at least Balances in HSA Prs m escor rip eti to ha nn dd ru oub g file lld s between 2014 and 2016.0 A .0 m 1ong enrol lees with e -m 0.p 0loy 8 ee-only coverage, balances 4.9 10% 4.7 5% Preventive Care in High-Deductible and Consumer-Directed Health Plans." American Journal of Managed Care 10% 0.13 Thi and s st the 5udy e ir fam xa ilim es. ine H s SA whe -eligib therle a nd heahow lth p H laSA ns b ar ae la e nxpe ces im cted p atc o t m us ae k e of he and alth p spe la nding n me m on be he rs m alth orc ea c re os serv t con icsc esiou . Most s be a ca cc uo se unt B untin, et al. (2006) found that accounts may offset the reduction in spending from HDHPs by 2–7 percent, but based 0.12 $500 7% $3,000. While somewhat arbitrary, this value represented the highest prevailing annual deductible in the sample, a level inc reased from $1,131 at the beginning of 2014 to $2,861 at the beginning of 2016 (Fig 3% ure 4). Similarly, balances more Report Availabili Blooty d te : Thi st, s r pae np eor l t is available on the internet a0 t .w 03 ww.ebri.or *** g 0.09 3% 222-230. 2% hold theye a rr s w 0.1 e p 0ill a ybing uildm up orea n ou H t SA of p boc alakn ecte t h (tao n in thet r daedgit riona ee tha l he t c aont lth p ribu lans tions . In a the re or my a, dhe e) a ble th capus lae n me most m b peerop s w leill a us re ehe leass lth y and the conclusion on studies published between 1996 and 2002, which was prior to the introduction of H 2% SAs. Conclusion and Limitations 4% above w 5% hich individuals may be more likely to c2% onsume discretionary care. than doubled among enrollees with family coverage. The average balance went from $1,416 in 2014 to $3,312 at the EKG, echocardiogram, stress test 0.01 0.02 un d o ne not ce us ssa er y a c lot ar e of ahe nd aw lth ill cbaer em serv ore ic cos es in a t cons ny c ious give n ye about ar .t he Ov he er a tim lth ca e, g rr eow serv ing icH es t SAh b aa t la the ncy e s m usea . y mitigate the impact of This study was conducted through the EBRI Center for Research on Health Benefits Innovation (EBRI CRHBI), with the At the outset of our analysis, we expected that larger HSA balances might result in increased use of health care B be unt ginn in, M ing eof lissa 20 16 Be.e uw It sh keould s, C he com ryl D e a as no mber su g, rp Arm ise elia tha Hta tvhe ila nd, per c Keant nik ag ae K of apa uc r, coun Nicole ts w Lit uh rie a, ze Rola ro nd bala Mc nc D ee a vt it tt, he and beM. ginning 00.00 X-ray 0.03 *** 0.11 *** H ESA ndno balanc tees s may affect use of health care services independently from contributions for a number of reasons. the deduc 0% 0% 0% $0tible. We find that as individuals build up balances in HSAs, they use more health care services than they fun Also ding include supp dor in t t of het he ana follow lyses ing weror e g the aniz fol atlow ioning s: Am on em He bw erit c t, ha Brlue act e Crris oss tic s: Blue ag e Shie , geld ndA ess r, oc ge ia og tion, raphic ICUB reA g, ion, JP Mor annu ga an l salary, 2007 2008 Blood Tes 20 t/Pa 09nel 2010 2011 EKG, Ech 20o 12 cardiogram 20 , 13 or Stress T20 est 14 2015 2016 X-Ray 2017 2018 services, effectively offsetting the effect of the high deductible. And indeed, we found evidence that higher HSA of the yeSus ar fe an M ll to ane rquis ar . ze 2r00 o a 6. " fteC r on twsu o y m ee arr-s of Direc con ted tr H ibut eaion lth C s.a Only re: E 2 p arlye E rcveide nt nc of eH A Sb Aou s stt a Er ff te ec dt s 20 On 16 C w ost ith A and zer Qua o ba lit la ync ." e, Less Than $1,00 E E Em m mp p ploy loy loye e ee e e- - -On On Only ly ly $1,000 –$2,999 $3,000 –$5,999 Fam Fam Family ily ily $6,000 or More CT scan 0.01 *** 0.03 *** ot Cont herrw ibut iseion would s are . I cn e app ss ee dnc by e ,a H nnu SAa b l st ala anc tue tor s m y lim ay ib ts. I lunn co t thent crost ast-, re bd auc lan ing ces b effuil ecd t of up high ove-rd te im duc e. tL ible arg he er a blt ah p lanla cens s m ov ae yr a Cn ind hase,ic Milli atorm for an, indiv Pfize idu r, a al nd vs. P hRMA family. coverage, and the number of covered dependents. Finally, to control for health balances resulted in increased use of health care services and higher spending. However, while the increases in use of regardless He of altw h he Aff the airrs tw he 51 e6nr -w olle 53e 0. ha d employee-only coverage or family coverage (Figure 5). At the beginning of 1 Source: MEBRI RI estimates from administrative data. 0.005 * 0.02 induce use of health care services to a larger degree than contributions. Most account holders will build up an account time. The implication is that employers may want to consider the impact of the size of the plan’s deductible. Another See http://devenir.com/wp-content/uploads/2018-Year-End-Devenir-HSA-Research-Report-Executive-Summary.pdf Source: EBRI calculations from the National Health Interview Survey. status, the Charlson Comorbidity Index (CCI), which is widely used in the extant literature as a gauge of general health health care services were statistically significant, the magnitudes were quite small. 20 14, 20 percent of those with employee-only coverage and 17 percent of those with family coverage had a zero In 2003, the M Me ud sc ic u alro es k Pe re le sc ta ripti l su on rge D rr yug, Improvement, and 0 Mod .004 ernization *** Act (MMA) 0 .inc 01luded a* provision that a pproach is t Souo rce: explore EBRI estim ed ate uc s a fro tion a m admrin ound istrativ te he data. long-term benefits of saving the HSA balance for health care expenses in balance (to So So Sotu u u he rce: rce: rce: d EBRI EBRI EBRI egrest est est eeim im im tha ate ate atet s s s fro fro fro cont m m m ad ad ad ribu m m min in int istrativ istrativ istrativ ions e e e ad d drata. ata. ata. e made) because most people are healthy and do not use a lot of health 2 status, was derived annually from medical claims for every member covered under the policy (Charlson, et al., 1987) C harlson, M. E., P. Pompei, K. L. Ales, and C. R. MacKenzie. 1987. "A New Method of Classifying Prognostic Comorbidity See Buchmueller (2009), Bundorf (2012), Buntin, et al. (2011), Fronstin and Roebuck (2013), and Fronstin, Sepulveda, and balance. In contrast, the percentage of accounts with balances between $1,500 and $2,999, and $3,000 or more nearly 4 created Health savings account (HSA)-eligible health plans. HSA-eligible health plans have deductibles that are retirement. care services in any given year. Ultimately, large HSA balances may offset the effects of the high deductible. Tabl One are e ao tf o C explore onteis nt re sl a ted to the size of the deductible. In large part, employers have taken a “set it and forget it” Fi (Dnd eyo, ing Che s frro kinm , aF nd ixe Ciol d ,Ef 19 fe 92 ct)s ( Qu Rea gr n, ee sts a io l.,n 20 Equ 05)a . tFor ion fa s milies, we retained the maximum score among all members in Longitudinal Studies: Development and Validation." Journal of Chronic Disease 40 (5): 373-83. Roebuck (2013b). tripled between the beginning of 2014 and 2016 among those with employee-only coverage (Figure 6). Among those higher than what individuals are used to seeing in more traditional types of health insurance. In 2019, HSA-eligible Statistical significance denoted as follows: ***p<0.01; **p<0.05; *p<0.10. I ant ttit rod ude uc w tion ith th ................................ e deductible level in ................................ HSA-eligible health p................................ lans. Between 2007 and ................................ 2018, deductibles in P .......................... POs have 4 linked to the policyholder. The health insurance eligibility and claims data for this study, which come from an employer with approximately Fi w 3 it gh fa ure m 13ily sh cow ove s t rahe ge , re the sult p s f errcom ent a the ge m of ua ltc iv ca oun riatte s w fixiteh b d ea ff la enc cts r es b egertew ssi ee on n $ m 3,00 ode0 ls a of nd H$ S5 A,999 bala m ncoers on e thaus n d e oub of he led a, lth whil cae re The purpose of this Issue Brief is to examine whether and how HSA balances impact use of and spending on health health plans were required to have deductibles of at least $1,350 for employee-only coverage and $2,700 for family See http://devenir.com/wp-content/uploads/2018-Year-End-Devenir-HSA-Research-Report-Executive-Summary.pdf Deyo, R. A., D. C. Cherkin, and M. A. Ciol. 1992. "Adapting a Clinical Comorbidity Index for Use with ICD-9-CM increased an average of 8.3 percent per year, while they have only increased an average of 2.5 percent in HSA-eligible 120,000 employees, were matched to HSA balance, contribution, distribution, and investment data from the EBRI HSA serv Dataic ................................ es. Regression analysi ................................ s is needed to sort out ................................ the effects from aging ................................ and other variables f ................................ rom the effect of HSA ..... 6 the percentage of accounts with balances of $6,000 or more quadrupled between the beginning of 2014 and 2016 care services. We expect balances will affect discretiona Fir gy u serv re 14 ices more than non-discretionary services. If higher HSA coverage. In addition, maximum out-of-pocket amounts cannot exceed $6,750 for employee-only coverage and 5 4 Administrative Databases." Journal of Clinical Epidemiology 45 (6): 613-9. Methods D he aa ta ltb h p ase. lan A s. su Ib t set of is possaible bout th 6,50 at in0 p trod olic ucy ing hold lae rr gs w er H as u SA-sed elig in t iblehe he st alt udy. h pla D n atdae d fruc om tible 20 14 op– t20 ions 16 ov we err e ti m ana e m lyze ayd c . ause In 2015, 15 percent of the population accounted for 67 percent of health care spending among the commercially insured Figur Figur Figur Figur e 11 e 9 e 2 e 6 ( bFi ala gur ncee s on 7). use of health care services. Only the key coefficients from the regressions related to the impact of HSA balances are associated w Liit nh hig ear Fhe ixe r d us Ee ff of ecthe s M alt o h c dea l rE est se im rv ait ce es s, ove of HS rA t iB m ae la , n the ce s do en d uc Sp te ibn le d iin t nghe HSA-eligible health plan $13,500 for family coverage. Certain primary preventive services — typically those deemed to prevent the onset of Variables ............................................................................................................................................................... 7 individuals to enroll in those options. To the degree that HSA balances were not large enough to cover the newer, population. In other words, 85 percent of the population accounted for only 33 percent of health care spending. See The research objective of the present study was to estimate the effect of HSA balance on use of health care services balance are presented. Othe An r vD n aUse u r is ia al tribu ble D of Se s inc istion tribu lT ude le ot ct of HS tion al d He in t HS s, he alth A A As b Balanc m y T od Ca sets, ype er l e w e Se e of Cov 2006 ra et r a vic Begin ge – e ,e 2018 s gr, e ag n 2014 nd in e e, rg , 2014 g of Y –e 2016 oge – ra ar 2016 phic , region, income, individual Account owner contributions have been trending up as well. Among those with individual coverage, the average annual K we ill y b fin ecdom ings: e a less effective tool in engaging individuals in their use of health care services. disease $bil— lion scan be and often are exempt from the deductible and covered in full. Otherwise, all health care services Fr onstin, Paul. 2019. "Enrollment in HSA-Eligible Health Plans: Slow and Steady Growth Continued Into 2018." EBRI Me highe http thod s:r / /d w se w ................................ dw uc .he tible althc s, t ohose stinstit but ala e.nc o................................ rg es w /resould earc h/ not pub re lis cult ations in hig ................................ /hche ci-re r us see arc of h/he entry alth c /to ................................ a pr -e sp . enders-among-the ............................... -commercially-insured -7 and spending. As in most observational data analyses, inferring causality was challenging due to potential endogeneity c or ont farm ibut ily 60 ion in covercarg ee a, sed spousa froml c $9 ov 03 er a to ge$1 , c,0 hil 64 dr e bn e/ot twe he en 2014 r depende and nt20 s, a 16 nd (Fi the gu rC eh 8 a)r. lson C Simila orm lyor , a bm idit ong y I nde those x. H wSA ith f ba am lail ny c es Policyholders With Employ Figur eee 4 -Only Coverage, 2014 –2016 must be Iss su ue b je Bc rie t tf, no. o the 47 pla 8n (E dm edpuc loy tible ee .B enefit Resea20 20 rc 14 14 h Ins 20 20t 15 15 itute 20 20 ). 16 16 Model 1 Model 2 ? In general, we find that larger beginning-of-year HSA balances result in increased use of health care services. increased-spending-concentration-and-consistent-turnover-from-2013-to-2015 or confounding. A diagram can be helpful in uncovering reasons why a naïve analysis might lead to biased estimates in w Rob ere us m tne eass su C rehe d c tw kso ................................ ways: in continuous (................................ thousands of) dollars a ................................ nd as an indicator for ................................ balances of at least $3 ................ ,000. The 8 coverage, the average annual contribution increased from $2,096 to $2,479. Not all contributions are being saved for $312 ,500 .0 Beginning of Year Balance, 53.8 Another approach to explore relates to education around 2014 the 20 H 15 long SA 20 B -t16 a elra m n cb ee nefits of the H SH ASA Ba . lH an Sc Ae balances can be saved Use of the following services increased: emergency department visits, primary care physicians (PCP) and 5 a specific context. Data Fr t rhe eEB The ons su R fut ltI ts in t MMA a in, P u ca re lc.ula he aN ul. 20 ls ot tions to w only o in18 e c fro lude q . d ua m "o Tr td d p ion a e e a top nds a s w p in r le ov e in Hea http tris a ek ion th e la s :rd /g lt /is w e h S ta ly w rtibut w c a aom v .llow kiion ngs ffp .o s indiv a s for rg rA a/c b he cle oun c a idu u w lth rr it ta - e h a c B ls nt oa s e la ts or fe nr n / w re c olle pe a p e s, C so xc d trt/ he e in HSA ont 2018 pa tions lt ribut h ca -e -.mp e ions r ligib eloy e, xpe le eD r -is he he ns traa ibut e lt ls, b th h p -ion but la ene ns s, a afi v te ts nd o r-a op sg urve In e e v n a a ennu yst /nd ma efun l nts, d 2011? 11.0 Finding50 s ................................ % ................................ by Type of Cov ................................ er(ag $th eo , u 2014 sands) –2016 ................................ ( = $ 3 , 0 0 0)................................ 9 for health care expenses in retirement on a tax-preferred basis. However, few individuals are doing so, and even fewer specialist visits, chiropractic services, physical therapy, psychotherapy, blood tests, X-ray, CT scan, MRI, and 50 HSAs, w 20 hi17 ch a : Erset im tax a- te es xeFr mo pm t ttrhe us tE or BRI cus HtSA od iD al a ata cb cou ase nt ." s f Eu Bnde RI Id ss w ue it h cont Brief no. ribut 46 ions 3 ( Eam nd ploy ass ee et s t Beh ne atfit indiv Resea iduracls h can d Thi ists st ribu udy m tions h ak ae vs us e bee e n tr of 6 te wnding o dat ahighe sourrc e as s w —e tll. he B E eB tw RI e eH n S20 A D 14 at a and base 20a 1n 6, a d he ve arla th insu ge ann ra ua ncl d e ia st nd ribut meion dics inc al cla re im ased s da fr ta om 6 10.2 are invesm ting usc tulos he Deir p k e H en le SA d ta e .l n su E t d V ruc g ae ra ira tyion b . l ecould result in more efficient use of HSAs. See https://www.ebri.org/health/publications/fast-facts/content/a-possible-disconnect-between-perception-and-utilization-of- To ad$3 $3 dr , ,5 0e 00 00 ss the potential endogeneity of HSA balance in analyses of use of health care services and $2,894 spending, we 2014 2015 2016 Fi Trg eur nds e 3 d in HSA epictss t ................................ he relationships betw ................................ een HSA balance and ................................ use of health care serv ................................ ices over two yea45.2 rs....................... By 9 In general, we found that larger beginning-of-year HSA balances resulted in increased use of health care services. Use 45% $3,312 10.0 $6 us 97 e tto o $9 p Ins a6 yt 3 a it for utm e he ) ong . alt h c accaoun re etxpe hold ns ee rs. s w Indivi ith em dua ploy ls w eeit-h HS only A cs b oveerne agfit e afr nd om in a c rtera iple sed t a frx a om d v $2 an ,384 tage t:o e$2 mp ,894 loye e a mong those from a large employer with employees residing throughout the United States. $2,710 hsas employed lineaIr n p fix ae tid e n et ffh eo ctssp m ita od l eling. This econometric techni -$4q 6ue .01 controls fo r all tim -$1 e5 -inv 7.3a 5ria 9.1 nt c haracteristics (both construction, HSA balances are affected by four mechanisms: 1) employer contributions to the HSA on behalf of of the following o serv Indivi iced s in uals cr e wait sed h $3, in one 000 o or r m bor ote h mod in the eir ls : HeSm Ae a rtg e the nc yb d eg ep inn aring tmeof nt tv he isi ty s, p earr im visi atre yd c P ar CeP s a phy nd sicspe ians c ia (P liC stP s ) Trceont nds ribut in U ion se s t of o H te he alt a h cc Cou are nt Se arre v ic de es d uc ................................ tible from taxable inc ................................ ome, any interest or ................................ other capital earnings on .................... assets in 11 with family coverage (Figure 9). 8.7 40% 38% observed and unobserved), removing their potential confounding effect. In the context of the present analysis, inclusion 40 Emergency department $5.20 $49.50 ** employees, which would increase balances; 2) employee contributions to his or her HSA, which would also increase a Fr nd ons spe $3 tin, P ,0c 00 iali ast ul, a visi nd ts, c M. hiCrhr opis ra top ctihe c serv r Roe ice bs, p uckhy . 20 si14 cal th a. "eBrraapnd y, -p Nsy am cho e a the nd ra Gp eyne , b ric lood Pr etsc est ris, X ption D -rayr, ug CT sc Usea n, AftMRI er A, da op nd tion of The E $2 B,RI 500 HSA Daon taba avse era isg e a one repr-e ha sent lf vais tiv it em ro erp eosi pe tor r yye of ar inf tha or n ind mativ ioidu n, d ae lsv e w $2,384 lop ith HSA ed to b fa ac la ilintc ae te s unde analyrses $3,on 000. the state of, $2,861 37.0 the HSA build up tax free, and distributions for qualified medical expenses from the HSA are excluded from taxable Findings from Fixed Effects Regression Equations ...................................................................................................14 35% of person 8.0-specific fixed effects eliminates underlying health status (yellow triangle) — as long as it is persistent over the balances; 3) dis Pt rr iibut mario y ns ca rte h a pt h t yhe sic ie am np vloy isits ee elects to take to p $a 5y .1 for 6 or reim *** burse the $m 19 selves .52 for c ** urrent or past out- m Wus hilc eulos d 35 isa % tkr e Fu ibu lelltt- a ions Re l su p la rh gc a ee vrm e y .e b H nt eo e, w n inc Ceon vesu rre , a m inpa sieng, r-tD ie ir ant n in ec ser tecdrv eH ic ae e sia s, ng lt h P pr neum la scn Wit rb ipti er oof h a n d a c rH ug ce oun afil lth Sa ls t hold , and vings A errs ha adiolog cvce oun b ye r tee ." n inv la E te Bd RI e st t o N ing. ot cae r d sThe io (E vm ap sc p elo ul rcy a ee r nt e age and individual behavior in, HSAs. The HSA database contained 5.9 million accounts with total assets of $13.4 billion as income to the employee. o Spending was $19.52 higher for primary care office visits and $28.06 higher for specialist visits. three-year study period — as a source of endogeneity (i.e., it breaks the links through the two solid blue lines). Time- of Re-su poc lts of ket he Rob altus Sh c pte ne a cr iss e al, i s w Ct he hic vic sh k its sw ................................ ould decrease balanc................................ es; and 4)$ c 5a .p 14 ital appr................................ ** eciation or $ d2 e8 p.r 0e 6ciatio ................................ n th **at may occur if the 16 disease (B Ee Kne G, fit e cRe hosea card rciog h Irns am tit , ust te re ) ss 35 (3 test)). were not affected by HSA balances. of of D aceccount . 31, hold 2017 er . s w Incit lude h inv d e in t stm he ent ds inc atase rt e a issed de tfr aile om d , 6 p yee t rd ce e-nt ide to nt13 ifie d p, erpce ern son t a- m leong vel info th 30.2 ose rm a wtit ion o h em n ind ployiv ee idua -only l a nd coverage 29% $2 $2, ,5 000 00 Overall spending was $496.40 higher for every $1,000 in HSA balances, though the finding was not $2,358 30% 30 varying health (e.g., acute conditions and other health shocks), however, can still lead to biased results. a ccount holder has elected to invest their savings. Chiropractic visits $2.75 * $19.96 *** between 2014 and 27%2016 (Figure 10). The percentage investing also doubled among those with family coverage. employer contributions, beginning- and end-of-year account balances, annual distributions, investments, and account- Conclusion and Limitations ................................................................ 26% ....................................................................16 statistically significant. Spending associated with other outpatient services was $550.96 higher among 6.0 25% 24.2 The Frons m tin, P agnitaude ul, a of nd the M. e C 24% ff hr eis cttsop whe asr not Roe la buc rge k.. I 20 n th 14e b . e"qQua uatilit on y w of he Hreea lt wh C e ea xa rem A ine fted r A wd he op the ting r H aSA Fu b lla -Re lanpcla ec s of eme ant t le , a Hst ig h- 25% Physical therapy $4.05 $19.95 owner demographics. $1,965 individuals with an HSA balance of at least $3,000 and was statistically significant. 4.9 The other problematic relationship that exists between HSA balance and use of health care services and spending is Refere $2 $1nc , ,0 500 00 es ...........................................................................................................................................................17 In Figure 3, the relationships of interest are the red arrows. Namely, the goal is to obtain an unbiased estimate of the $3,000 aD ffe ed ctuc ed t ible use H of eahe lth P alth ca lan Wi re tserv h a H ice ea s, t lth Sa he la vings A rgest e cff coun ect tw : a As Fi re vla e-tY ed e atro St out udy." patie Ent BR vIis Iiss ts ( ue a B cr om ief, no bina. tion o 404 f primary Muc Trend h re ssea in rU ch ha se of s b H ee en d altone h C a to reunde Servic rstae nd s the impact that the deductible in HSA-eligible health plans, and high- Psychotherapy $1.17 -$23.87 19.3 19% 20 20% depicted in Figure 3 as the dotted blue arrow from health services use in Year 2 and employee contribution at the direct effect (ß; assumed to be constant over time) of HSA balance on use of health care services and 18%2spending. ? Balances in HSAs more than doubled between 2014 and 2016. Among enrollees with employee-only coverage, care and (spe Emp cia loy list ee v B isi etne s)fit . Ou Re r sea find rc ings sugg h Instituteest ). that individuals with $3,000 or more in their HSA at the beginning of the Health insurance eligibility and medical claims data from an employer with a 17% pproximately 120,000 employees were Endnotes .............................................................................................................................................................18 deductible health plans (HDHPs) more generally, has had on use of health care services and spending. The initial 4.0 Other outpatient costs $133.53 *** $550.96 *** 15.5 Commensurate with higher beginning-of-year balances, use of health care services has been increasing among HSA- beginning of the same year. This refers to the p $963 ossibility that individuals who anticipate future health care consumption Plaguing the analysis, however, are the other relationships that exist between these two variables. Namely, use of year visitb ea dla PnC cP es a s inc nd respe ased cia fr lis otm s on $1,a13 ve1 a ragte t he one b-e ha ginn lf ving isit of mor 20 e14 pe tro y$2 ear ,861 th $1,416 a n ind at the iv idua begin ls nin witg h H of SA 20 16 ba.l aSim nceils unde arly, r matche $1 $1, ,d 5 000 00 to the EBRI HSA Database. Because the employer offered a choice of health plans, including health studies t 15e %nded to focus on broad questions like who enrolls in an HDHP, how enrollees differ from non-enrollees, risk $850 12.2 Prescription drug -$34.54 -$161.04 eligible health plan enrollees. Between 2014 and 2016, there was an increase in the number of inpatient hospital Fronstin, Paul, and M. Christopher Roebuck. 2013. "Health Care Spending after Adopting a Full-Replacement, High- may be more likely to contribute to their HSA. In this case, it is not HSA balance that is leading to health services use health care services and costs in each year are determined by underlying health status — which may or may not be balances more t $697 han doubled among e 9.9 nrollees with family coverage. The average balance was $3,312 at the $3,000. maintenance organizations (HMOs), preferred provider organizations (PPOs), exclusive provider organizations (EPOs), selection, and the impac $1,131 t of HDHPs on overall use of services and spending. Recent studies examined more targeted Blood test, panel -$0.89 9% -$1.28 9% admissi10 ons, inpatient hospital days, specialist visits, and prescription drug fills (Figure 11). There was also an increase Figures 2.0 10% 7.2 1.8 b ut the op De pd osit ucteible : us H e eof althe h P ala lth ca n Wirteh serv a He ica els is th Sa aff vings A ecting cH coun SA b t: alA a nc Fiv ee s.- YUnf earo St rtuna udy." te ly E,B r ReIv I ess rse uec a Bus riea f, no. lity is 38 diff 8 icult to beginning of 2016. observable—and will depend on both chronic and acute conditions. Further 1.7 more, two indirect relationships (depicted as 2.0 1.6 and HSA-eligible health plans, a subset of about 6,500 policyholders was used in the study. This sample is smaller than questions, such as medication a 5.5 dherence for individuals with chronic conditions (Fronstin, Sepulveda, and Roebuck, $1 $5 ,000 00 EKG, echocardiogram, stress test $1.72 $8.00 in laboratory services and imaging (Figure 12). Not all health care services experienced an increase in use. There were 1.1 Figure 14 ( Esh mow ploy s t ehe e B re 3.4 ene sufit lts f Re rom sea r the ch I m ns utltitiv ua te ria ).t e fixed effects regression models of HSA balances on overall spending both assess and correct for, but we undertook several exercises to delve deeper into this issue. soli Figur d e b lue 1, H lin SA es) -E lm igib ight le H aleso c althonfound Plan Enrtoll hem aena ntly , si 20 s. 0 7 Fi–r20 st, 1p 8r ................................ ior use of health care ................................ services may result in e ........................ mployees 6 the total number of enrollees in the HSA-eligible health plan in any given year for a number of reasons. Individuals 5% 0.7 2013a); gener 1.7 ic drug use (Fronstin and Roebuck, 2014a); whether individuals with an HDHP are likely to price shop ? The percentage of accounts with a zero balance at the beginning of the year fell to nearly nothing after two X-ray $4.57 *** $8.99 no changes or mixed findings related to primary care office visits, CT scans, MRI, chiropractic visits, physical therapy, 0.5 and spending by type of health care service. It shows the commensurate increase in spending associated with the increasing their subsequent contributions in order to later reimburse themselves (causing an increase in HSA balance). were included only if they were continuously enrolled in the HSA-eligible health plan over the 2014–2016 period. For (Brot-Gold 0 berg, et al., 2015); quality of health care received (Fronstin and Roebuck, 2014b); use of specialty years of contributions. Only 2 percent of HSAs started 2016 with a zero balance. At the beginning of 2014, 20 Figure 2, Total HSA CT sc aA nssets, 2006–2018 ................................................................ $2.41 ................................ $16.89 ................... 6 Fr psy ons cho tin, P thera aul, M py, aa nd rtin J. musc Se ulo pulv ske ele da ta , l s and urgM. eryC h (not rist op shhe owrn i Roe n fig buc ur ke . s) 20 . 13 These a. "C c onsu hange me s m r-Dair ye b ctee d d rH ive ea n b lth P y tla he ns si Re mp dle uc fa e c The t Robust $5 0% 0.0 00 $0 ness Checks increase in use of health care services shown in Figure 13. For instance, spending was $19.52 higher for primary care Second, distr20 ibu 06tions 20 m 07ay be20 ta 08ken to p 2009ay for 20 p 10rior us 20e 11 of he20 alt 12h care 20 13 service 20 s ( 14causin 20g 15 a dec 20 r16 ease in HSA 2017 ba 20 la 18 nce). each year, the datasets were matched by gender, birth year, zip code (5 digit), health plan deductible, employer percent of those with employee-only coverage and 17 percent of those with family coverage had a zero medications (Fronstin and Roebuck, 2018); delays in breast cancer screening (Wharam, et al., 2019); and the Inpatient Le Ho ss sp Tha ital Ad n $500 missions E (p mp eloy r ee-On Inply atient $500 Hosp – ita $1,499 l Days (per 100 Sp $1,500 ecialis –$2,999 t Visits Family Pre$3 scri ,0 p00 o tionr Dru Mog Fi re lls that as this Long pop -Te M ula R rm Ition Use a g Of O es the utp ya us tiee n tm Por hy esi he cia an Vi lth csi atrs A e ser nd vP icre e$ s. sc 7 .r 6 ip 7tion Drugs." *** Health Af $35.f0 a1 irs 32 (6) **: 1126-1134. office visits and $28.06 higher for specialist visits. Overall spending was $496.40 higher, though the finding was not Stated differently, si 10 n0 en ce ro us llee e sof ) health care serv eic nro ells i eess c ) orrelated within individual over time, a naïve (cross-sectional) W e undertook several auxiliary analyses to evaluate the robustness of our results. As mentioned above, reverse c Fiont gur re ibut 3, ion Rela to tion the sh H ips SA, Be at nd wea en ind n HSA ic aB to ala r for nce e a m nd ploy He ea elt -on h Se ly rv vs. ic e fs Use amily ................................ coverage. A unique, e................................ ncrypted ID was used...... to 9 combinatb ion o alanc f d e.e T dhe uc tp ible erc e sint zea, gp er e of senc acceoun of ta s w n HSA ith b or ala H nR ce As b , and etwte ye pn $1,50 e of HD0 a HP nd (Ha $2 vil,99 and, 9, a end t al.$,3,00 2011 0 o b)r. m Re ocre ent ne sa tr udie ly s Musculoskeletal surgery $4.98 $83.10 * statistically significant. Spending associated with other outpatient services was $550.96 higher among individuals with analysis will yield a negative estimate for ß because individuals who use less health care (i.e., healthy) will tend to have causality$0 is a key threat. Since foreseeable health services utilization is arguably more likely to be for chronic conditions Source: http://devenir.com/wp-content/uploads/2018-Year-End-Devenir-HSA-Research-Report-Executive-Summary.pdf track people over time as well. tripledEBRI Is betwe sue en th Briefe i s re beg giinn ste reing d in of the 20 U.S 14 . P aa te nd nt a20 nd 16 Tra d aem mark ong O ftfhose ice. ISS w Nit : h e 0887 m –1 p3loy 7Xe /9e 0- 0only 887 – 1c3ov 7Xe /9 r0 a g $ e .5.0 A +.m 50ong those with a Fr re ons als tin, P o exa aul, M mining artm in J. oreSe cu prulv rent ed d aa , ta and and M. pC la hn e ristnr op olle here s’ Roe bebha ucvkior . 20 s ove 13br. long "Med eic r a titm ion U e petrili iod zas. tion a Whil nd e tA he dhe re ra ernc e e in a So Sou urce: rce: EBRI EBRI est estim imate ates s fro from m ad ad Em m m ploy in inistrativ istrativ ee-Only e e d data. ata. Family Figure 4, BSo eg urce: inn O v ing EBRI eraof est ll im s Y p e ate e an r s d fro Bia n m la g ad nc me in, istrativ by Ty e d pata. e of Coverage, 20$ 14 96– .2016 90 ................................ $496.40 ................................ ... 10 an HSA balance of at least $3,000 and was statistically significant. This category captures many services that were not higher HSA balances, and individuals who use a lot of health care (i.e., unhealthy) will tend to have lower HSA (i.e., acute health shocks are far less predictable), we re-estimated our models on the subset of our sample (about two- famil© y c 2o 0v 1e 9r,a E gm e, plo the yee B perc ene ent fit ag Resea e of a rch I ccou ns ntts w itute ith b –Educa alant cio es b n ae ntd wResea een $1,50 rch Fund. 0 and All $2,99 rights 9 mor reserv e th ed. an d oubled, differencH es in ealth S resu av lting s as A cross ccoun thet -st Eudie ligible s, t Phe lan." consi Am st ee ric nta n J find our ing na isl of thaMa t Hna DH ge Pd s r C eadruc e 19 e us (12 e of ): he e4a 00 lth ca -e40r7. e services. easily classi Sfied ource: int EBRI o estimate the spe s c frifi om c ad ser min vistr ice ative s md eata. ntioned, and the largest component of that category was not specific balances. Indeed, we estimated pooled cross-sectional ordinary least squares (OLS) models and found negative and thirds) for which the Charlson Comorbidity Index was equal to zero throughout the entire study period. By removing while the percentage of accounts with balances of $3,000 or more quadrupled between the beginning of 2014 Figure 5, Percentage of HSAs With Zero Balance at Beginning of Year, by Type of Coverage, 2014–2016 .................. 11 Statistical significance denoted as follows: ***p<0.01; **p<0.05; *p<0.10. enough to know what the service was for. indiv iduals and families with chronic conditions, we reduced the likelihood for reverse causality to be an alternate and 2016. e e e e A r e e e e e e e e e e e e eb b b b b b b b b b b b b b b b br r r r r r r r r r r r r r r r re i. i. i. i. i. i. i. i. i. i. i. i. i. i. i. i. i.sea o o o o o o o o o o o o o o o o or r r r r r r r r r r r r r r r rg g g g g g g g g g g g g g g g grch IIIIIIIIIIIIIIIIIs s s s s s s s s s s s s s s s ss s s s s s s s s s s s s s s s s re u u u u u u u u u u u u u u u u ue e e e e e e e e e e e e e e e epo B B B B B B B B B B B B B B B B Brt r r r r r r r r r r r r r r r r rief ief ief ief ief ief ief ief ief ief ief ief ief ief ief ief ief fro • • • • • • • • • • • • • • • • • m M M M M M M M M M M M M M M M M M t a a a a a a a a a a a a a a a a ah y y y y y y y y y y y y y y y y y e 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 E 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3,,,,,,,,,,,,,,,,,B 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 R0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 I E 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 19 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9du • • • • • • • • • • • • • • • • • cat N N N N N N N N N N N N N N N N No o o o o o o o o o o o o o o o oi.................o 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 n8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 an 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 d Research Fund © 2 019 Employee Benefit Research Institute 16 17 14 10 11 13 15 12 18 7 6 5 3 2 8 9 4

Do Accumulating HSA Balances Affect Use of Health Care Services and Spending?

Do Accumulating HSA Balances Affect Use of Health Care Services and Spending?

Volume 482

Pages 18

EBRI_IB_482_RoleofHSAs.23May19

EBRI Issue Brief

May 23, 2019

Paul Fronstin

M. Christopher Roebuck

Health