"Primary Care By Dionne S. Kringos, Wienke Boerma, Jouke van der Zee, and Peter Groenewegen doi: 10.1377/hlthaff.2012.1242 HEALTH AFFAIRS 32, NO. 4 (2013): 686–694 Europe’s Strong Primary Care ©2013 Project HOPE— The People-to-People Health Foundation, Inc. Systems Are Linked To Better Population Health But Also To Higher Health Spending Dionne S. Kringos (d.s [email protected]) is a ABSTRACT Strong primary care systems are often viewed as the bedrock of postdoctoral health systems health care systems that provide high-quality care, but the evidence researcher in the Department of Social Medicine at the supporting this view is somewhat limited.We analyzed comparative Academic Medical Center, primary care data collected in 2009–10 as part of a European Union– University of Amsterdam, in the Netherlands. funded project, the Primary Health Care Activity Monitor for Europe. Our analysis showed that strong primary care was associated with better Wienke Boerma is a senior researcher at NIVEL, the population health; lower rates of unnecessary hospitalizations; and Netherlands Institute for relatively lower socioeconomic inequality, as measured by an indicator Health Services Research, in Utrecht. linking education levels to self-rated health. Overall health expenditures were higher in countries with stronger primary care structures, perhaps Jouke van der Zee is a part- time professor in the because maintaining strong primary care structures is costly and Department of International promotes developments such as decentralization of services delivery. Health at Maastricht University, in the Netherlands. Comprehensive primary care was also associated with slower growth in health care spending. More research is needed to explore these Peter Groenewegen is the director of NIVEL. associations further, even as the evidence grows that strong primary care in Europe is conducive to reaching important health system goals. rimary care is the first level of pro- health of populations and the performance of fessional care, where people pre- health systems, and they suggest directions for sent their health problems and further research. wheremosttherapeuticandpreven- In Europe these studies have evoked an in- 1 P tive health needs can be satisfied. creased interest in the great variation among Strong primary care is believed to contribute health systems and the different roles assumed tohigh-performinghealthcaresystems,abelief by primary care. The question that we believed 1–4 that is supported by evidence to some extent. neededtobeansweredwaswhetherresultsfrom Decisionmakershavetrustedthisevidenceand previousstudiesaboutthebenefitsofstrongpri- invested in primary care reforms, such as the marycaresystemswouldstillbevalidusingmore AffordableCareActintheUnitedStates,aswell recent data and more tailor-made measures. asinnumerouschartersandstatementsmadeby Also,wewondered,couldtheresultsbegeneral- 5,6 nongovernmental organizations worldwide. izable if many more European countries were Several studies that compareprimary carein- considered? ternationally and within the United States have In 2009–10, as part of a European Union– provided evidence of the benefits of strong pri- funded project, the Primary Health Care mary care, in terms of better opportunities to Activity Monitor for Europe, we performed a control costs, improved quality of care, better systemic literature review to derive seventy- population health, and less socioeconomic in- seven indicators. These measured five key di- 1–4 equality in health. These studies have shown mensionsofprimarycare:structure,access, co- the potential of primary care to improve the ordination,continuity,andcomprehensiveness. 686 Health Affairs April 2013 32:4 Downloaded from content.healthaffairs.org by Health Affairs on April 17, 2013 by Rachel McCartneyWiththisapproach,thestudyaimedtocoverthe secondary care only upon referral by a primary 2 complexityofprimarycarebyaddressingitasa careprofessional.DianaDelnoijandcolleagues 7,8 multidimensional concept. showedthathealthcaresystemsinwhichfamily Dataontheindicatorswerecollectedinthirty- physicians served as gatekeepers to more spe- one countries from the international literature, cializedcarehadalowerincreaseinambulatory governmental publications, statistical data- care costs and in the use of outpatient health bases, and national expert consultations. To services but not in total health care costs, com- quantify the strength of the five primary care paredtohealthcaresystemswithdirectlyacces- 12 dimensions in a country, the data on each indi- sible specialist care. cator were transformed into a score ranging Fromthesestudieswecaninferthatthegate- 9 from 1 (weak) to 3 (strong), inspired by keepingfunction,usuallycoupledwithpatients’ 3 James Macinko and colleagues’ approach (see being registered with a primary care doctor, 10 online Appendix 1). seemstobeakeyelementleadingtolowerhealth We tested the relationship in thirty-one spending. However, patients do not express European countries between the strength of equalsatisfactionwithallaspectsofprimarycare the five primary care dimensions, on the one when gatekeeping is present. 13 hand, and key health care system performance Madelon Kroneman and colleagues showed indicators,ontheotherhand:healthcarespend- that patients in countries with a gatekeeping ing,patient-perceivedqualityofcare,potentially system were less satisfied with the quality of avoidable hospitalizations, and population nonmedicalaspectsofprimarycare,suchascon- health and socioeconomic inequality. Specif- venience in obtaining an appointment or wait ically,wesoughtanswerstothefollowingques- timesintheofficebeforeseeingthedoctor,than tions and tested the associated hypotheses. patientsincountrieswithdirectlyaccessiblespe- First, is health care spending lower, and the cialists. However, differences in satisfaction increase in spending slower, in countries that with nonmedical aspects of access were not re- haverelativelystrongprimarycare,afteradjust- lated to patients’ ratings of the quality of the ing for national income? actual care received, such as quick relief of Second, is the patient-perceived quality of symptoms. nonmedical aspects of primary care lower in Otherstudies—mainlyfromtheUnitedStates, countries that have relatively strong primary where gatekeeping has limited implementa- care? Recently published research seems to in- tion—revealedotherproblemswithaccesstopri- dicate that this relationship exists. marycare.Thesestudiesconfirmedthatlimited Third, are potentially avoidable hospitaliza- availability of primary care in the United States 14–16 tions lower in countries that have relatively increased avoidable hospitalizations. A hos- strongprimarycare,afteradjustmentfordisease pital admission is potentially avoidable when it prevalenceandtheavailabilityofhospitalbeds? couldhavebeenpreventedbyeffectiveoracces- 17 Fourth, is population health better in coun- sibleprimarycare. Hospitalizationofso-called tries that have relatively strong primary care, ambulatory care–sensitive conditions, such as after adjustment for risk factors? asthma and chronic obstructive pulmonary dis- Fifth,aresocioeconomicinequalitiesinhealth ease, are particularly avoidable if well managed smaller in countries that have relatively strong in primary care. primarycare,afteradjustmentforinequalitiesin Positiveassociationsbetweentheaccessibility risk factors? of primary care and better population health 18–20 Aftersomebackgroundinformationonrecent have been identified in literature reviews. developments in primary care research, we re- Studies, mostly undertaken in the United port our findings below. States, have shown that regions with a higher primarycarephysiciandensity,butnotahigher specialist density, have a healthier population Recent Research On The Effects Of than regions with a higher specialist and lower Primary Care primary care physician density as measured by Largeandincreasingproportionsofnationalin- total and cause-specific mortality, low birth- 18–20 comes are spent on health care. Data from the weight, and self-reported health. 1990s show that countries with strong primary Little evidence is available of a relationship care spent less and were better able than other between socioeconomic inequality in health countries to contain rising health care costs. and the strength of primary care. Several US 11 Ulf Gerdtham and colleagues found that studies suggest that access to primary care can theoverallcostofhealthcarewasgenerallylower reduce socioeconomic andracial inequalitiesin 20,21 incountrieswhereprimarycareperformsagate- health. However,thisresulthasnotyetbeen 22,23 keeper function and patients can thus access clearly confirmed in international studies. April 2013 32:4 Health Affairs 687 Downloaded from content.healthaffairs.org by Health Affairs on April 17, 2013 by Rachel McCartneyPrimary Care Study Data And Methods The control variable for diabetes was the per- Asnotedabove,weuseddatagatheredin2009– centageofobeseoroverweightpopulation(body 10 as part of the Primary Health Care Activity massindexof25orhigher),bysexandage(ages 9 29 Monitor for Europe. The database covered 15–54 and 55+). The control variable for both thirty-one European countries (twenty-seven ischemic heart disease and cerebrovascular dis- European Union member states as well as easewasage-andsex-standardizedhypertension 28 Switzerland, Turkey, Norway, and Iceland). prevalence. The control variable for chronic Depending on the availability of data, some asthma, bronchitis, and emphysema was data analysesweredoneforasmallernumberofcoun- on the self-reported smoking prevalence in the 10 26 tries. Appendix 2 contains the descriptive sta- populationagesfifteenandolder. Moredetails 10 tistics, a list of included countries per variable, are available in the online Appendix. and sources of all included variables. ?SOCIOECONOMIC INEQUALITY IN HEALTH: Variables And Confounders The level of socioeconomic inequality was mea- ?HEALTH CARE SPENDING: The total level of suredbythehighestattainededucationallevelin health carespending was measured by the total having poor or very poor self-perceived health healthcareexpenditurepercapita, inUS dollar status, asthma, and diabetes, by calculating an 24 purchasing power parity, in 2009. Its growth age- and sex-standardized Concentration Index 28 was measured over the period 2000–09, as foreachcountry. Thisindexquantifiesthede- 10 shown in Appendix 3a. gree of education-related inequality by condi- Thecontrolvariablesofthewealthandgrowth tion, ranging from 1 to -1. It indicates that a in the wealth of a country were measured by condition is more concentrated among people gross domestic product per capita in US dollar with a higher (when positive value) or lower purchasingpowerparityin2009andchangesin (when negative value) educational back- 24 30 gross domestic product during 2000–09. ground. Zero points indicates equality; see 10 ?PATIENT-PERCEIVED QUALITY: The patient- Appendix 3e. perceived quality of nonmedical aspects of pri- As control variables, the age- and sex- 25 mary care was measured by the age- and sex- standardized concentration index for obesity 27 standardized percentage of people who rated (relatedtodiabetes)anddailysmoking (related the quality of care received from family physi- toasthmaandself-perceivedhealth)wereused. cians as very good or fairly good, as opposed to IndependentVariables Dataonthestrength 25 10 fairly bad or very bad (see Appendix 3b). of primary care were derived from the Primary ?POTENTIALLY AVOIDABLE HOSPITALIZA- HealthCareActivityMonitorforEuropeproject 9 TIONS: Potentially avoidable hospitalization database. The following five independent was measured by age-standardized hospital variables were used for the strength of pri- admission rates per 100,000 population, by mary care: structure, accessibility, continuity, sex,forthreechronicdiseases:asthma,chronic coordination, and comprehensiveness. obstructive pulmonary disease (bronchitis and ?STRUCTURE: Thefirstvariableindicateshow emphysema), and diabetes (short-term compli- primary care in a country has been structured. 26 cations). Elements are the existence of primary care pol- Control variables were the prevalence of dia- iciesandregulations—forexample,onequaldis- 27 betes, asthma,andchronicbronchitis/emphy- tributionofprimarycareprovidersandfacilities; 28 sema, age-standardized, by sex and total; and theavailabilityoffinancialresourcesforprimary the total number of available hospital beds per care;thepopulation’scoverageforprimarycare 24 100,000 population. services; and the development of workforce for ?POPULATION HEALTH: Population health primary care—for example, workload, age, and 7,8 was measured by potential years of life lost, by training of family physicians. sex, due to diabetes; ischemic heart disease Because these aspects of primary care struc- (heart disease characterized by reduced blood ture are positively associated with each other, flow to the heart muscle, often related to coro- theirsummationresultsinonevariableindicat- nary artery disease and hypertension); cerebro- ing the overall strength of a country’s primary 9 vasculardisease(stroke);andobstructedairway care structure. conditions, including bronchitis, asthma, and ?ACCESSIBILITY: Theremainingvariablesre- 26 emphysema. flect the strength of important aspects of the 7,8 Potential years of life lost is a summary mea- primary care services delivery process. The sureof premature—thatis,preventable—deaths accessibility of primary care was measured by that weighs deaths occurring at younger ages the national and geographic supply of primary more highly than those occurring at later ages, careservices;thewayaccessisorganizedinpri- age-standardized per 100,000 population (ages marycarepractices—forexample,theuseofap- 0–69). pointmentsystemsandtheorganizationofafter- 688 Health Affairs April 2013 32:4 Downloaded from content.healthaffairs.org by Health Affairs on April 17, 2013 by Rachel McCartneyhourscare;andtheaffordabilityandacceptabil- reliability of the sources varied across the ity of services as perceived by patients. thirty-one countries. Also, thirty-one countries ?CONTINUITY: Continuityofprimarycarewas is a relatively small number from a statistical measuredbyconditionsinplaceforanenduring pointofview.Someanalysescouldbeperformed doctor-patient relationship—for example, pa- onlyforevenfewercountries,becauseoflimited tients’ being registered with a primary care data availability. The number of included coun- doctor;provisionsinplacetoestablishinforma- tries rangedfrom thirty-one countries (for thir- tionalcontinuityofcare—forexample,theuseof teen out of fifty-five variables) to twenty coun- electronicclinicalrecordsystems;andaspectsof tries (for the diabetes admission rate per 10 the quality of the doctor-patient relationship— 100,000 population). Appendix 2 contains a forexample,patient-perceivedavailableconsul- listofincludedcountriespervariable.Asaresult, tation time. wewerenotabletoincludetheimpactofpoten- ?COORDINATION: Coordination of primary tiallyimportantcontextfactors—suchasculture, care was measured by the existence of a gate- politics,andhealthcaresystemtype—onthede- keeping system, the skill mix of primary care pendentvariables.Itisrecommendedthatfuture providers,thecollaborationwithinprimarycare studies take this into account. and with secondary care providers, and the in- Another limitation is that some of the data tegration of certain public health functions in have been collected at the national level, but primary care. disaggregateddatawould haveallowed analysis ?COMPREHENSIVENESS: Thecomprehensive- intointracountryvariation.Thisstudyshouldbe ness of primary care was measured by the usedasastartingpointformorein-depthstudies breadthofservicesofferedtopatientsatthepri- on each of the complex outcome areas, prefer- mary carelevel—forexample,medicaltechnical ably by also using microlevel data. procedures and certain preventive services. 10 Appendix 1 provides an overview of all indica- tors used for each of the dimensions. Study Results Dependent Variables Becausetheseprocess Strong primary care was associated, respec- functionswerenotstronglyassociatedwitheach tively,withhigherlevelsofhealthcarespending, other, four dependent variables were used: pri- but also a reduced rate of growth in health care mary care access, coordination, continuity, and spending; lower rates of potentially avoidable 8 comprehensivenessofprimarycare. Allfivede- hospitalization; better population health out- pendent variables were continuous, ranging comes; and lower socioeconomic inequality in from 1(relativelyweak)to3 (relativelystrong). self-rated health (see Exhibit 2 and Appendices 10 Exhibit 1 provides an overview of the resulting 3 and 4). primarycarescoresbycountry,usingthescoring Total Health Care Expenditures Total 10 system shown in Appendix 1. health care expenditures were higher in coun- Statistical Analyses The association be- trieswithastrongerprimarycarestructureafter tween dependent and independent variables adjustment for national income than in coun- was evaluated in simple (Pearson correlation) tries with a weaker primary care structure and multivariable regression analyses. In the (Exhibit2).However,countrieswithmorecom- simple linear regression analyses, only one de- prehensive primary care services delivery had pendent and one independent variable were slower growth in total health care expenditures used. In the multivariable analysis, one control percapita, also after adjustment for the growth variable was added (to prevent overdetermina- innationalincome(therateofchangeis-0.20; 10 tion).Bothtypesofanalysiswereperformedfor see Appendices 3 and 4). Patient-perceived allhypothesesbyusingeachofthefiveprimary quality of nonmedical aspects of care and the carestrengthmeasuresasindependentvariables strength of primary care were not associated inseparateanalyses.WeusedthesoftwareSPSS/ with any aspect of strong primary care. PASW Statistics, version 18.0. Potentially Avoidable Hospitalizations Strengths And Limitations A strength of Stronger primary care structure is associated this study is that it demonstrates the contribu- withlowerhospitaladmissionratesforasthma, tionofprimarycaretotheperformanceofhealth for both the total population (reduction rate: care systems at a European level. The study has -0.45) and males (reduction rate:-0.51). measured the complexity of primary care in di- Countries with more comprehensive primary versehealthcaresystemsusingacomprehensive carealsohadlowerhospitaladmissionratesfor set of indicators. However, the strength of pri- asthmacomparedtocountrieswithlesscompre- marycarewasmeasuredatonemomentintime. hensive primary care,both for thetotal popula- A limitation of the study is that although the tionandforwomen(reductionrates:-0.36and best available information was used, the -0.37, respectively). These lower rates were April 2013 32:4 Health Affairs 689 Downloaded from content.healthaffairs.org by Health Affairs on April 17, 2013 by Rachel McCartneyPrimary Care Exhibit 1 Strength Of Key Primary Care Aspects In Thirty-One European Countries, 2009–10 Country Structure Accessibility Continuity Coordination Comprehensiveness Austria 2.22 2.27 2.19 1.38 2.33 Belgium 2.21 2.13 2.38 1.70 2.53 Bulgaria 2.14 2.15 2.33 1.44 2.54 Cyprus 1.91 2.11 2.32 1.49 2.19 Czech Rep. 2.14 2.35 2.41 1.64 2.33 Denmark 2.38 2.46 2.43 1.96 2.40 Estonia 2.29 2.21 2.42 1.71 2.41 Finland 2.31 2.20 2.32 1.74 2.51 France 2.16 2.06 2.33 1.63 2.47 Germany 2.20 2.25 2.38 1.38 2.34 Greece 2.10 2.08 2.25 1.96 2.17 Hungary 2.08 2.34 2.33 1.46 2.29 Iceland 1.77 2.28 2.40 1.60 2.42 Ireland 2.20 1.96 2.38 1.57 2.36 Italy 2.33 2.27 2.31 1.73 2.13 Latvia 2.14 2.15 2.38 1.65 2.41 Lithuania 2.27 2.29 2.30 1.98 2.56 Luxembourg 1.90 2.03 2.31 1.63 2.42 Malta 2.12 2.17 2.17 1.82 2.38 Netherlands 2.50 2.38 2.26 2.20 2.32 Norway 2.27 2.25 2.36 1.56 2.55 Poland 2.12 2.35 2.33 1.92 2.29 Portugal 2.41 2.34 2.35 1.62 2.47 Romania 2.31 2.26 2.33 1.55 2.20 Slovak Rep. 2.02 2.27 2.39 1.39 1.98 Slovenia 2.36 2.47 2.30 1.84 2.32 Spain 2.43 2.44 2.43 1.84 2.51 Sweden 2.23 2.17 2.25 2.32 2.49 Switzerland 2.04 2.17 2.37 1.63 2.42 Turkey 2.27 2.05 2.15 1.61 2.36 UK 2.52 2.40 2.37 1.88 2.52 SOURCEKringosDS.ThestrengthofprimarycareinEurope(Note9intext).NOTEScoresrangefrom1(weakprimarycare)to3(strong primary care). partly caused by the difference in hospital bed Furthermore, the comprehensiveness of pri- supply amongcountries, since lower admission mary care was also positively associated with a rateswereassociatedwithhavingfewerhospital reduction in potential deaths due to cerebro- beds.Chronicobstructivepulmonarydiseasead- vascular disease among the total and male pop- missionratesofmenwerealsolowerincountries ulations (reduction rates: -0.42 and -0.43, re- withastrongercoordinationofprimarycare(see spectively).Thisassociationwaspartlycausedby 10 Appendices 3 and 4). variation in the prevalence of hypertension Countries with better access to primary care among the respective groups. were associated with lower hospital admission However, when one takes into account the rates for diabetes, for both the total population prevalence of hypertension, there is a strong (reduction rate: -0.40) and males (reduction association between primary care structure 10 rate:-0.46; see Appendices 3 and 4). and fewer potential deaths due to cerebro- Population Health Countries with stronger vascular disease among men (reduction rate: 10 primary care structures were associated with -0.36; see Appendices 3 and 4). fewerpotentialdeathsduetoischemicheartdis- Both the structure and the coordination of ease among the total, male, and female popula- primary care were associated with fewer poten- tions. Countries’ having more comprehensive tial deaths due to chronic asthma, bronchitis, primary care was also associated with fewer po- and emphysema. Countries with a stronger tential deaths due to ischemic heart disease structure of primary care were associated with amongmen(thereductionrateinpotentialyears fewerpotentialdeathsamongwomenduetoob- 10 oflifelostwas-0.35;seeAppendices3and4). structive airway conditions (reduction rate: 690 Health Affairs April 2013 32:4 Downloaded from content.healthaffairs.org by Health Affairs on April 17, 2013 by Rachel McCartney"