The Health and Well-being of an ACO Population

April 9, 2019
Thomas E. Kottke, MD, MSPH

,
Jason M. Gallagher, MBA

,
Marcia Lowry, MS

,
Sachin Rauri, MS

,
Juliana O. Tillema, MPA

,
Jeanette Y. Ziegenfuss, PhD

,
Nicolaas P. Pronk, PhD, MA

,
Susan M. Knudson, MA

Volume 25, Issue 4

Among HealthPartners plan members, musculoskeletal, psychosocial, and neurologic conditions create the greatest burden to current health; diet offers the greatest opportunity to improve future health scores; and 42% report a high level of well-being.

ABSTRACT

Objectives: To identify opportunities to improve the health and well-being of members of HealthPartners, a health plan based in Minnesota.

Study Design: Cross-sectional analysis of insurance claims, death records, and survey data.

Methods: We calculated a current health score from insurance claims and death records for all 754,584 members 18 years and older who met inclusion and exclusion criteria for the period January 1, 2015, to December 31, 2015, and/or January 1, 2016, to December 31, 2016. Adjusting responses to represent the member population, we calculated a future health score based on 7 items and a 1-item well-being score from survey data that we collected between July 1, 2015, and December 31, 2016.

Results: Forty-four percent of the loss to the current health score among HealthPartners members is attributable to musculoskeletal, psychosocial, and neurologic conditions. Among the 7 components of the future health score, the greatest opportunity for improvement (31% of the total potential) is increasing dietary fruits and vegetables. Although 42% of the members reported high levels of well-being, 14% reported low levels. On average, members with the lowest levels of well-being were insured by a Medicaid product and had low educational achievement.

Conclusions: By applying the summary measures of health and well-being to the HealthPartners member population, we identified opportunities to address conditions that created a high burden on current health, opportunities to improve prospects for future health, and subpopulations who would benefit from interventions that would increase their sense of well-being.

Am J Manag Care. 2019;25(4):182-188Takeaway Points

We assessed HealthPartners members’ health and well-being for the period July 1, 2015, to December 31, 2016, and found that:

  • Three broad classes of conditions—musculoskeletal, psychosocial, and neurologic conditions—are responsible for 44% of the loss in the current health score.
  • Increasing healthy eating is the greatest opportunity to improve the future health score.
  • Although 42% of the members reported a high level of well-being, 14% reported a low level of well-being.

These data provide a baseline and strategic guidance as HealthPartners works to improve the health and well-being of its members, patients, and community.Although valid measures of health and well-being have been proposed, tested, and implemented by various organizations, HealthPartners has been unable to find a feasible set of measures that would meet its needs as a health plan striving to improve its performance on the population health component of the Triple Aim.1 Therefore, the health plan developed its own summary measures of health and well-being for adults that are composed of 3 scores: a current health score, a future health score, and a well-being score (Table 1). The current health score is the complement of disability-adjusted life-years (DALYs) and is composed of the disease burden due to mortality before age 75 years and morbidity.2,3 The future health score has 2 components. One is based on 6 member-reported behaviors (physical activity, eating fruits and vegetables, tobacco use, alcohol use, adequate sleep, and healthy thinking). The other is an age- and sex-specific preventive services score that is based on the performance of the medical group to which the health plan attributes the member. The elements of this measure are consistent with the Healthcare Effectiveness Data and Information Set 2016 preventive services measures. The well-being score is based on a single question that is used by the Organization for Economic Co-operation and Development and is considered valid by experts in the field of subjective well-being: “How satisfied are you with your life?”4,5 Because HealthPartners can link survey responses to claims data, the summary measures of health and well-being allow it to measure its performance in greater detail than is possible with publicly reported measures like the County Health Rankings or proprietary measures like the Gallup-Sharecare Well-Being Index.6,7 HealthPartners will use the measures to guide its strategic planning of health and well-being initiatives and to track its progress on the population health component of the Triple Aim.

METHODS

The HealthPartners Member Population

We conducted the analyses that we report here to improve HealthPartners’ operations, quality management, and member care and experience. They are based on 754,584 health plan members who met the following criteria: (1) continuously enrolled (with allowance for a 30-day gap in insurance coverage or discontinuance of enrollment due to death); (2) enrolled during the period January 1, 2015, to December 31, 2015, and/or January 1, 2016, to December 31, 2016; and (3) 18 years or older without upper limit. Among the members meeting these criteria, we excluded (1) members known to prefer a language that is not English; (2) members with an address of a correctional facility, nursing home, or hospice; and (3) the 0.2% of members whose names appear on the health plan, research, or related do-not-contact lists.

We based the current health score on all member insurance claims and death records for the defined period. We based the future health score and well-being score on responses to the member survey. Each month, we asked a stratified random sample of members who met the inclusion and exclusion criteria described above to complete a mail survey. We contacted nonresponders by telephone. We designed the survey sample to account for differential nonresponse (ie, by gender, age, and type of insurance product) that was present in a pilot survey of a similar population. We oversampled Medicaid members so that we would have an adequate number of responses to compare with those of members insured through commercial products or Medicare. We asked the survey respondents to self-identify their ethnicity, race, and education. However, we only report non-Hispanic white versus all other categories at this time because the number of respondents is not large enough to make more detailed ethnicity or race comparisons meaningful.

Over the period of interest, the survey response rate averaged 39.0%. We based our current analysis on 9144 survey responses and insurance claims of 754,584 individuals. We reweighted the survey responses to reconstruct the member population that met our inclusion and exclusion criteria. The attributes of the reconstructed population fall within 0.3 percentage points of the actual population, except for insurance product, for which the difference is 1 percentage point.

Calculating DALYs and the Current Health Score

As we described in detail in another publication,8 the HealthPartners current health score is composed of the burden of morbidity (years lost to disability [YLDs]) and mortality (years of life lost [YLLs]) on the population. To identify the burden that specific conditions put on the population as morbidity (YLDs), we assigned Global Burden of Disease (GBD) condition weights9 to HealthPartners insurance claims grouped by the Johns Hopkins ACG [adjusted clinical group] System expanded diagnostic clusters (EDCs) and major expanded diagnostic clusters (MEDCs).10 YLDs for an individual can range from 0 for an individual who has no insurance claims during a year to 1. When a member’s claims tracked to more than 1 EDC within an MEDC, we attributed the disease burden to the EDC with the largest GBD weight, and we only counted 1 weight within an MEDC for a particular member. We used the method described by Kottke et al to calculate the DALYs due to morbidity for members who had claims in more than 1 MEDC so that a member could not have a YLD value greater than 1.11 Although the choice of the age used to calculate YLLs is arbitrary and the GBD collaborators use the age of 86 years for life expectancy,12 we elected to follow the convention of the County Health Rankings6 and calculate the mortality burden from death as YLLs before age 75 years. To calculate the DALYs attributable to an MEDC (eg, musculoskeletal), we summed all the YLDs and YLLs attributable to the MEDC.

We created the current health score for individuals by calculating the complement of that individual’s DALYs multiplied by 100. For example, there were 232,720 DALYs lost by the 754,584 individuals in the population, yielding an average DALY loss of 0.31. The current health score would be 69. We divided the DALYs attributable to an MEDC by the total DALYs for the population and multiplied by 100 to generate the loss to the current health score attributable to the MEDC.

Calculating the Future Health Score

The HealthPartners future health score has 2 components: a behavior component based on survey responses (the items and pass criteria are available in eAppendix A [eAppendices available at ajmc.com]) and a preventive services component. The potential range of the score is 0 to 7. Being most interested in knowing the number of individuals who might benefit from a particular behavior change intervention and wanting to keep the measures easy for the end user to understand, we gave each behavior a score of 1 if present and 0 if absent, rather than assigning a weight based on an association with an outcome variable.

The preventive services score is specific to the member’s age and sex, has a range of 0 to 1, and is based on their medical group’s performance on the 2015 adult primary care and obstetrics/gynecology preventive services score that HealthPartners reports in its annual clinical indicators report.13 However, not all medical groups see enough HealthPartners members to be scored, and some patients have not been seen by a medical group that has been scored. If we have attributed a patient to a medical group to which we have given a preventive services score, we gave the patient the same score as we gave the medical group. If we have not given a member’s attributed medical group a score, we gave the member the average of all medical groups with a score, without accounting for the member’s age or sex. Finally, if a member has not had a claim in 2 years, we gave them a score of 0.

We adopted this scoring system because some preventive services (eg, screening for colon cancer with colonoscopy) are indicated very infrequently and others (eg, screening mammography) depend on risk factors, like family history or member preference, that are not captured by claims data.

Calculating the Well-being Score

We used the member survey as the data source of the well-being score. The survey offers the respondent a response scale with a range of 0 to 10 to the question, “How satisfied are you with your life?” We arbitrarily defined a response of 9 or 10 as a high level of well-being, a response of 7 or 8 as a moderate level of well-being, and a response of 0 to 6 as a low level of well-being.

Statistical Methods

We tested the relationships among the 3 health score components and demographic variables of age, gender, race, education, and Medicaid coverage using approaches appropriate for the various types of distributions encountered in these metrics and covariates. We report the relationships among demographic variables and the future health score from a general linear model using SAS Proc Genmod (SAS Institute; Cary, North Carolina). We employed nonparametric methods to test relationships among well-being and the current health score with the other measures due to violation of normality assumptions. We used the Kruskal-Wallis test to report relationships with categorical variables, and we used Spearman correlations to test relationships with ordinal measures. We used logistic regression to evaluate the probability of being a nonuser—that is, having no insurance claims during the period of interest. We used bootstrapping14 to draw 1000 random samples of 300 members from the weighted data to estimate 95% CIs. We used SAS version 9.4 (SAS Institute; Cary, North Carolina) to conduct the analyses. The HealthPartners Institutional Review Board has determined that this analysis constitutes quality improvement.

RESULTS

Population Attributes

The HealthPartners member population is relatively young, non-Hispanic white, and educated (Table 2). Just 10% are 65 years or older. Slightly more than half are women, and the majority are non-Hispanic and white. Fewer than 15% of the members lack some college or technical school, and more than half have a college degree or an advanced degree. Nearly 90% of the members are insured through a commercial product; relatively similar proportions of the remaining members are insured through Medicare or a Medicaid program, including Minnesota Seniors Health Option.

Current health scores. The current health score declines significantly with age (P <.0001 for trend), and men have a higher score than women. Non-Hispanic white members have a lower current health score than members of other ethnic groups and races. The current health score rises with increasing education (P <.0001 for trend). Members insured through Medicare have the lowest current health score (P <.0001 vs other products).

When statistically adjusted for age, the difference in the current health score between men and women persists. When statistically adjusted for age and sex, differences in the current health score between non-Hispanic whites and the group of members of any other race or ethnicity do not persist.

Future health scores. Mean future health scores are significantly lower for members aged 50 to 64 years compared with the other age groups. The future health score is significantly greater for women than for men. Likewise, the trend for the future health score to increase with increasing education is statistically significant (P <.0001 for trend). Non-Hispanic white members have a higher future health score than members of other ethnic groups and races. Members insured through a Medicaid program have significantly lower future health scores than members insured through Medicare or commercial products. When statistically adjusted for sex, the difference in the future health score between non-Hispanic white members and the group of members of any other race or ethnicity persists (P = .0045).

Well-being scores. Although the well-being score does not differ significantly by age, sex, or race/ethnicity, the increase in well-being scores associated with increasing education is statistically significant (P <.0001). Members insured by a Medicaid program have significantly lower well-being scores compared with members insured through Medicare or a commercial insurance product.

Overall Summary Measure Score Values

The overall average current health score is 69 (95% CI, 62-73) for the 754,584 members as a group during this period (Figure). The musculoskeletal, psychosocial, and neurologic MEDCs are the leading causes of loss to the current health score. The future health score averages 4.45 (95% CI, 4.42-4.48) of a maximum possible score of 7.0. Increasing fruit and vegetable consumption and increasing healthy thinking present the greatest opportunities to improve the future health score. For the well-being score, 42% of the population reported high well-being, 44% reported moderate well-being, and 14% reported low well-being.

Relationships Between the Number of Members With a Claim and the Current Health Score

The correlation between the number of members with a claim associated with an MEDC and the DALYs associated with an MEDC was 0.52 (P <.006). The MEDCs with the greatest number of members with a claim were administrative, musculoskeletal, eye, and skin (Table 3). These 4 MEDCs accounted for 26.2% of the DALYs. Claims for preventive services contributed to the large number of claims in the administrative MEDC but did not generate DALYs. Musculoskeletal, psychosocial, and neurologic were the MEDCs that caused the most loss to the current health score—44.8% of all loss when taken together.

Associations Among the Health and Well-being Scores

Twelve percent of HealthPartners members have a current health score of 30 or less (Table 4); 34% of the DALYs are attributable to this group of members. At the other end of the spectrum, 34% of the members have a current health score that is greater than 90; their experience accounts for 2% of the DALYs. The remaining DALYS are fairly evenly distributed among the members who have a current health score between 31 and 80. Although the association of the overall future health score with the current health score is relatively weak (r = 0.021; P = .042) due to a statistically significant negative association of the preventive services component with the current health score (r = —0.177; P <.001), the association of the behavior component with the current health score is statistically significant while weakly positive (r = 0.050; P <.0001). The increase in the well-being score with an increasing current health score is nearly monotonic and is statistically significant (r = 0.093; P <.001).

Only 16% of the members have a future health score of 0 to 3 (eAppendix B Table 1). More than 75% of the members have a future health score in the range of greater than 3 to 5, and only 7% of the members have a future health score greater than 5 to 7. The well-being score increases monotonically by more than 1.5 points as the future health score increases (r = 0.296; P <.0001).

DISCUSSION

From our analysis of all 754,584 member claims, we found that 3 broad classes of conditions—musculoskeletal, psychosocial, and neurologic—are responsible for 44% of the loss in the current health score. Within these broad categories and in order of descending burden, the leading musculoskeletal conditions are generalized musculoskeletal conditions, lower back pain, and cervical spine pain; the leading psychosocial conditions are anxiety and depression; and the leading neurologic conditions are headache, migraine, and head injury. From survey responses given by a randomly selected sample of members, we found that increasing healthy eating, as indicated by whether the member meets the goal of eating 2 servings of fruit and 3 servings of vegetables every day, is the greatest opportunity to improve the future health score. Finally, we found that although 42% of the members reported a high level of well-being, 14% are apparently suffering as they reported a low level of well-being. These data provide a baseline as HealthPartners works to improve health and well-being in partnership with its members, patients, and community.

Our analysis also suggests that small changes in behavior might lead to a large improvement in current health. We calculated that a 1-point increase in the mean future health score could be associated with a reduction in 13,084 DALYs (eAppendix B), an amount that is greater than the total number of DALYs attributed to either respiratory or cardiovascular conditions. The information provided by the summary measures of health and well-being is revealed neither by analyzing the frequency of claims nor by analyzing the cost of care for particular conditions. It will provide significant guidance as HealthPartners develops its strategic initiatives to improve population health.

The well-being measure provides important information about the social determinants of well-being: Among HealthPartners members, education and the adequacy of financial resources (using Medicaid as a proxy) are strongly linked to well-being. These insights compel the organization to continue its many initiatives to increase health equity.15-17 If it were to rely on publicly available social indicators, HealthPartners would not have the same level of insight or be able to track its progress at a level of detail it would find sufficient.

Limitations

We acknowledge that HealthPartners’ summary measures of health and well-being have many limitations. The current health score, based on insurance claims, can be biased by both differential access to care and a member’s propensity to seek care. The future health score and the well-being score are both based on limited information. We acknowledge that more information in each of the scores would increase their accuracy, but we found that our initial, more extensive survey resulted in unacceptably low response rates that we could not increase by any feasible means. Even with the response rates that we are getting with a brief survey, bias is possible if the members who are suffering the most are also the ones who are least likely to respond. Better calibration would also increase our confidence in our scores. Finally, at this time, we do not have enough survey responses to analyze scores by ethnicity and race, but our capability will increase as we continue to survey our members.

CONCLUSIONS

We believe that, if applied widely, the HealthPartners measures could help guide national efforts to improve population health and well-being, and we invite others to test the validity and reliability of our methodology. If accountable care organizations and other healthcare organizations would collect, report, and use a uniform set of summary measures of health and well-being, they would not only be able to track their own progress in the populations that they serve, but they would also make a significant contribution to implementing the recommendations made in the 2011 report For the Public’s Health: The Role of Measurement in Action and Accountability.18 When aggregated, the measures could guide regional and national efforts to improve population health and well-being because nearly 90% of Americans are now registered with a health plan.19Author Affiliations: HealthPartners (TEK, JMG, ML, SR, JOT, NPP, SMK), Minneapolis, MN; HealthPartners Institute (TEK, JYZ, NPP), Minneapolis, MN.

Source of Funding: HealthPartners was the sole and exclusive source of funding.

Author Disclosures: The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (TEK, JMG, SR, JOT, JYZ, NPP); acquisition of data (JMG, ML, SR, JYZ, NPP); analysis and interpretation of data (TEK, JMG, ML, SR, JYZ, NPP); drafting of the manuscript (TEK, NPP); critical revision of the manuscript for important intellectual content (TEK, ML, SR, JOT, JYZ, NPP, SMK); statistical analysis (ML, SR); provision of patients or study materials (TEK); obtaining funding (TEK, NPP, SMK); administrative, technical, or logistic support (TEK, ML, SR, JOT, NPP, SMK); and supervision (TEK, JMG, SMK).

Address Correspondence to: Thomas E. Kottke, MD, MSPH, HealthPartners, 8170 33rd Ave S, MS 21110X, Minneapolis, MN 55425. Email: Thomas.e.kottke@healthpartners.com.REFERENCES

1. Berwick DM, Nolan TW, Whittington J. The triple aim: care, health, and cost. Health Aff (Millwood). 2008;27(3):759-769. doi: 10.1377/hlthaff.27.3.759.

2. Murray CJ, Lopez AD. Measuring the global burden of disease. N Engl J Med. 2013;369(5):448-457. doi: 10.1056/NEJMra1201534.

3. Metrics: disability-adjusted life year (DALY). World Health Organization website. who.int/healthinfo/global_burden_disease/metrics_daly/en. Accessed November 27, 2014.

4. Organisation for Economic Co-operation and Development. Output and analysis of subjective well-being measures. In: Organisation for Economic Co-operation and Development. OECD Guidelines on Measuring Subjective Well-being. Paris, France: OECD Publishing; 2013:179-247.

5. Diener E, Oishi S, Lucas RE. National accounts of subjective well-being. Am Psychol. 2015;70(3):234-242. doi: 10.1037/a0038899.

6. What and why we rank. County Health Rankings & Roadmaps website. countyhealthrankings.org/explore-health-rankings/what-and-why-we-rank. Accessed February 26, 2019.

7. Gallup-Sharecare Well-Being Index website. wellbeingindex.sharecare.com. Accessed February 26, 2019.

8. Kottke TE, Gallagher JM, Lowry M, et al. Validating a measure to assess disease burden from insurance claims. Am J Manag Care. 2019;25(2):e39-e44.

9. Salomon JA, Vos T, Hogan DR, et al. Common values in assessing health outcomes from disease and injury: disability weights measurement study for the Global Burden of Disease Study 2010 [erratum in Lancet. 2013;381(9867):628]. Lancet. 2012;380(9859):2129-2143. doi: 10.1016/S0140-6736(12)61680-8.

10. Concept: Adjusted Clinical Groups (ACG) - overview. University of Manitoba website. mchp-appserv.cpe.umanitoba.ca/viewConcept.php?printer=Y&conceptID=1304. Updated January 17, 2019. Accessed February 26, 2019.

11. Kottke TE, Faith DA, Jordan CO, Pronk NP, Thomas RJ, Capewell S. The comparative effectiveness of heart disease prevention and treatment strategies. Am J Prev Med. 2009;36(1):82-88. doi: 10.1016/j.amepre.2008.09.010.

12. Murray CJ, Atkinson C, Bhalla K, et al; U.S. Burden of Disease Collaborators. The state of US health, 1990-2010: burden of diseases, injuries, and risk factors. JAMA. 2013;310(6):591-608. doi: 10.1001/jama.2013.13805.

13. Bussey S, Kopischke R. 2018 Clinical Indicators Report: 2017/2018 Results. Minneapolis, MN: HealthPartners; 2018. healthpartners.com/ucm/groups/public/@hp/@public/documents/documents/entry_203585.pdf. Accessed February 26, 2019.

14. Miller DP. Bootstrap 101: obtain robust confidence intervals for any statistic. Presented at: 29th Annual SAS Users Group International Conference; May 9-12, 2004; Montreal, Canada. Paper 193-29. support.sas.com/resources/papers/proceedings/proceedings/sugi29/193-29.pdf.

15. Children’s Health Initiative. HealthPartners website. healthpartners.com/hp/about/community-engagement/childrens-health-initiative/index.html. Accessed February 26, 2019.

16. Kottke TE, Pronk N, Zinkel AR, Isham GJ. Philanthropy and beyond: creating shared value to promote well-being for individuals in their communities. Perm J. 2017;21:16-188. doi: 10.7812/TPP/16-188.

17. Isham GJ, Zimmerman DJ, Kindig DA, Hornseth GW. HealthPartners adopts community business model to deepen focus on nonclinical factors of health outcomes. Health Aff (Millwood). 2013;32(8):1446-1452. doi: 10.1377/hlthaff.2011.0567.

18. Institute of Medicine. For the Public’s Health: The Role of Measurement in Action and Accountability. Washington, DC: The National Academies Press; 2011.

19. Key facts about the uninsured population. Kaiser Family Foundation website. files.kff.org/attachment//fact-sheet-key-facts-about-the-uninsured-population. Published December 7, 2018. Accessed February 26, 2019.