Advertisement

Demonstrating Subpopulation Analytics

A Paradigm Shift for Improving Population Health
Open AccessPublished:August 05, 2015DOI:https://doi.org/10.1016/j.amepre.2015.05.028

      Introduction

      Population health outcomes are distributed disproportionately among the many subpopulations composing an aggregated total population.
      Middlesex-London Health Unit
      Evidence-informed program interventions are most likely to be applied effectively when target subpopulations are defined with optimal precision.
      • Dicenso A.
      • Bayley L.
      • Haynes R.B.
      Accessing pre-appraised evidence: fine tuning the 5S model into a 6S model.
      Comprehensive reviews of current approaches for defining and measuring population health find that they are typically composed of lists of county-level indicators, which lack the granular data to measure subpopulation disparities and provide little or no synergy for priority setting or the formulation of intervention strategies at the local level.
      National Quality Forum (Commissioned Paper)
      An environmental scan of integrated approaches for defining and measuring total population health by the clinical system, the government public health system and stakeholder organizations. Authors.
      • Shah S.N.
      • Russo E.T.
      • Earl T.R.
      • Kuo T.
      Measuring and monitoring progress toward health equity: local challenges for public health.
      We use Florida mortality data to demonstrate a new approach to subpopulation analytics that stages multiple sources of event-level data in a conformed data warehouse, with an online analytic processing interface and customized algorithms providing flexible and powerful analytic capabilities.

      Methods

      We used individual mortality data from the Florida Department of Heath for every death during the 3 years of 2008–2010, analyzed in 2014. Death (mortality) was the outcome of interest and it was characterized by the following variables:
      • 1.
        gender (male, female);
      • 2.
        race (black, white, other);
      • 3.
        age band (0–17, 18–44, 45–64, ≥65 years);
      • 4.
        cause of death (National Center for Health Statistics X 113);
      • 5.
        year (X 3); and
      • 6.
        county (X 67).
      This combination of variables yielded 545,112 subpopulations statewide or 8,136 per county. If we replaced the county with the ZIP code as the level of geography, the number of subpopulations grew to 7.3 million. To simplify the demonstration, we show only the subpopulations from a single Florida county (Orange) that have ten or more deaths and compare blacks and whites for the combined 3-year period.
      Four comparative measures were used to rank-order the subpopulations:
      • 1.
        number of deaths;
      • 2.
        death rate per 100,000 population;
      • 3.
        Z-score; and
      • 4.
        difference between black and white Z-scores (racial disparity).
      The Z-score normalizes the mortality rates to a common standard across the range of causes of death and age bands by computing the state-based mean rate and subpopulation SD for each age-banded cause of death, identifying those subpopulations at the outer boundary of the population distribution.
      We show only the top 25 subpopulations ranked by each single comparative measure.

      Results

      Each of the comparative measures produced a very different view of the Orange County subpopulations. Those furthest from the state mean death rates (Z-score) were mostly black, homicide by firearms; neoplasms of the prostate, uterus, breast, colon; HIV; and diabetes (Table 1). The subpopulations sorted by the size of the racial disparity, by contrast, were more equally divided by race. The white causes of death with disparities favoring blacks included melanoma; accidental poisoning and exposure to toxic substances (including drug-related deaths); and suicides (Table 2). Only a few of the subpopulations with the highest death counts appeared in the top 25 for either the Z-score or racial disparity rankings. Fixed groups of county-level indicators and their static logic models are incapable of analyzing the complex interactions of determinants and active agents from which local health status emerges.
      • Wolfson M.C.
      Notes on measurement and accountability.
      Table 1Top 25 Subpopulations by Z-Score and Total Deaths (Black and White)
      Cause of deathGenderRaceAgeDeathsRateZ-score
      Ranked by Z-score (black and white)
       Assault (homicide) by discharge of firearmsMaleBlack18–448664.882.40
       Assault (homicide) by discharge of firearmsMaleBlack45–641117.642.37
       Malignant neoplasm of prostateMaleBlack45–641219.242.23
       Malignant neoplasms of corpus uteri and uterus, part unspecifiedFemaleBlack≥651659.392.20
       Malignant neoplasm of breastFemaleBlack18–441510.972.18
       Certain conditions originating in the perinatal periodMaleBlack18–44118.32.13
       Human immunodeficiency virus (HIV) diseaseMaleBlack45–643962.532.11
       Diabetes mellitusMaleBlack18–44118.32.1
       Atherosclerotic cardiovascular disease, so describedMaleBlack45–643048.12.02
       Other diseases of arteries, arterioles, and capillariesFemaleBlack≥651244.552.01
       Malignant neoplasm of prostateMaleBlack≥6545235.531.94
       Certain conditions originating in the perinatal periodMaleBlack0–175147.621.86
       Essential (primary) hypertension and hypertensive renal diseaseFemaleBlack≥652696.511.81
       Cerebrovascular diseasesMaleBlack45–643048.11.79
       Renal failureFemaleBlack45–641724.771.79
       Malignant neoplasms of colon, rectum, and anusMaleBlack45–642438.481.76
       Diabetes mellitusFemaleBlack45–643652.461.75
       Congenital malformations, deformations, and chromosomal abnormalitiesFemaleBlack0–171413.741.73
       Certain other intestinal infectionsMaleWhite≥652925.911.68
       Human immunodeficiency virus (HIV) diseaseMaleBlack18–442619.611.67
       Hypertensive heart diseaseMaleBlack45–641930.461.66
       Hypertensive heart diseaseFemaleBlack≥652696.511.66
       Malignant neoplasm of ovaryFemaleWhite≥657349.751.66
       Malignant neoplasm of esophagusMaleWhite≥655650.041.59
       All other and unspecified malignant neoplasmsMaleWhite18–44173.991.54
      Ranked by number of deaths (black)
       All other diseasesFemaleBlack≥65139515.980.55
       All other forms of chronic ischemic heart diseaseFemaleBlack≥65108400.910.04
       Assault (homicide) by discharge of firearmsMaleBlack18–448664.882.40
       Cerebrovascular diseasesFemaleBlack≥6570259.850.61
       Malignant neoplasms of trachea, bronchus, and lungMaleBlack≥6564334.970.66
       All other forms of chronic ischemic heart diseaseMaleBlack≥6562324.51–0.31
       Certain conditions originating in the perinatal periodMaleBlack0–175147.621.86
       Cerebrovascular diseasesMaleBlack≥6551266.930.68
       Diabetes mellitusFemaleBlack≥6549181.891.09
       Alzheimer diseaseFemaleBlack≥6548178.180.62
       All other forms of heart diseaseFemaleBlack≥6547174.470.11
       Malignant neoplasm of prostateMaleBlack≥6545235.531.94
       Diabetes mellitusMaleBlack≥6541214.591.52
       Certain conditions originating in the perinatal periodFemaleBlack0–174039.251.42
       All other forms of chronic ischemic heart diseaseMaleBlack45–644064.131.15
       Other chronic lower respiratory diseasesMaleBlack≥6540209.36–0.10
       Malignant neoplasms of trachea, bronchus, and lungFemaleBlack≥6540148.48–0.63
       Human immunodeficiency virus (HIV) diseaseMaleBlack45–643962.532.11
       All other forms of heart diseaseMaleBlack≥6538198.890.38
       Malignant neoplasms of trachea, bronchus, and lungMaleBlack45–643759.320.24
       Diabetes mellitusFemaleBlack45–643652.461.75
       Renal failureFemaleBlack≥6534126.211.20
       Acute myocardial infarctionFemaleBlack≥6534126.21–0.52
       Other chronic lower respiratory diseasesFemaleBlack≥6532118.79–0.77
       Malignant neoplasm of breastFemaleBlack45–643145.171.46
      Ranked by number of deaths (white)
       All other diseasesFemaleWhite≥65857584.100.87
       All other forms of chronic ischemic heart diseaseMaleWhite≥65646577.250.86
       All other forms of chronic ischemic heart diseaseFemaleWhite≥65625425.980.16
       Other chronic lower respiratory diseasesFemaleWhite≥65431293.750.53
       Malignant neoplasms of trachea, bronchus, and lungMaleWhite≥65407363.690.86
       Cerebrovascular diseasesFemaleWhite≥65383261.040.62
       Other chronic lower respiratory diseasesMaleWhite≥65368328.840.79
       Malignant neoplasms of trachea, bronchus, and lungFemaleWhite≥65360245.360.04
       Alzheimer diseaseFemaleWhite≥65317216.061.08
       All other forms of heart diseaseFemaleWhite≥65292199.020.38
       Acute myocardial infarctionFemaleWhite≥65245166.98–0.08
       All other forms of heart diseaseMaleWhite≥65240214.460.55
       Acute myocardial infarctionMaleWhite≥65240214.460.43
       Cerebrovascular diseasesMaleWhite≥65198176.93–0.19
       Symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classifiedFemaleWhite≥65192130.860.22
       Malignant neoplasms of trachea, bronchus and lungMaleWhite45–6418164.020.40
       Diabetes mellitusFemaleWhite≥65179122.000.31
       Malignant neoplasm of prostateMaleWhite≥65178159.061.13
       Malignant neoplasm of breastFemaleWhite≥65170115.871.41
       All other forms of chronic ischemic heart diseaseMaleWhite45–6416558.360.92
       Accidental poisoning and exposure to noxious substancesMaleWhite18–4415636.641.32
       Diabetes mellitusMaleWhite≥65156139.400.54
       Alzheimer diseaseMaleWhite≥65144128.670.02
       Motor vehicle accidentsMaleWhite18–4413030.541.34
       Heart failureFemaleWhite≥6513088.600.87
      Table 2Top 25 Subpopulations by Racial Disparity
      Selected raceOpposite race
      GenderRaceAgeCause of deathZ differenceDeathsRateZ-scoreDeathsRateZ-score
      MaleWhite45–64Falls2.65248.491.480–1.17
      MaleWhite≥65Malignant neoplasms of kidney and renal pelvis2.593833.961.320–1.27
      MaleWhite≥65Malignant melanoma of skin2.452724.131.230–1.22
      MaleBlack45–64Assault (homicide) by discharge of firearms2.431117.642.3751.77–0.06
      MaleWhite≥65Intentional self-harm (suicide) by discharge of firearms2.323127.71.380–0.94
      MaleBlack18–44Assault (homicide) by discharge of firearms2.298664.882.404310.100.11
      MaleWhite45–64Malignant melanoma of skin2.25217.431.180–1.07
      MaleWhite18–44Accidental poisoning and exposure to noxious substances2.1615636.641.3275.28–0.84
      MaleBlack18–44Diabetes mellitus2.15118.32.181.88–0.05
      MaleWhite45–64Other and unspecified nontransport accidents and their sequelae2.13144.950.770–1.36
      MaleBlack45–64Human immunodeficiency virus (HIV) disease2.073962.532.113713.090.04
      MaleWhite≥65Certain other intestinal infections2.032925.911.68210.47–0.35
      FemaleBlack45–64Diabetes mellitus2.023652.461.755017.10–0.27
      MaleBlack45–64Cerebrovascular diseases1.993048.11.794315.21–0.20
      MaleBlack18–44Certain conditions originating in the perinatal period1.95118.32.1392.110.18
      MaleWhite≥65Malignant neoplasms of meninges, brain, and other parts of central nervous system1.932724.131.0415.23–0.89
      MaleBlack45–64Hypertensive heart disease1.931930.461.66269.20–0.27
      FemaleBlack45–64Renal failure1.911724.771.79175.82–0.12
      FemaleBlack18–44Human immunodeficiency virus (HIV) disease1.862518.291.51112.73–0.35
      FemaleWhite≥65Atherosclerosis1.793020.450.6513.71–1.14
      FemaleBlack≥65Other diseases of arteries, arterioles, and capillaries1.761244.552.012517.040.25
      FemaleBlack0–17Congenital malformations, deformations, and chromosomal abnormalities1.751413.741.73135.79–0.02
      MaleWhite≥65Parkinson disease1.729887.571.34631.4–0.38
      MaleBlack0–17Certain conditions originating in the perinatal period1.725147.621.863514.870.14
      MaleBlack45–64Malignant neoplasm of prostate1.701219.242.23207.070.53

      Discussion

      As demonstrated by our simple example using only a single data source and outcome, understanding community health status requires the ability to define and analyze subpopulations in multiple ways and for many specific outcomes. Varying the selection and weighting of prioritizing criteria will significantly change the rankings of community health outcomes.
      • Studnicki J.
      • Fisher J.W.
      Determining community health status priorities in an online analytic (OLAP) environment.
      Existing sources of publicly available granular de-identified data and powerful information technology enable hundreds of millions of subpopulations to be defined and analyzed for multiple outcomes, such as deaths (as in our demonstration), avoidable hospitalizations, inpatient and ambulatory surgical complications, readmissions, newly diagnosed cancers, emergency room visits, and many others. Because of the event-level nature of this data, integration with clinical data from electronic health records and other treatment subpopulations is feasible. Subpopulation analytics that is deployed and evolved on a national scale will inform population health improvement in the same way that genomics and metabolomics are informing clinical disease management.

      Acknowledgments

      This methods development and demonstration study were partially supported by a grant from the Charlotte Research Institute awarded to Dr. Studnicki. The Charlotte Research Institute had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication. Drs. Studnicki and Fisher had full access to all study data and take responsibility for the integrity of the data and accuracy of the data analysis. Study concept and design: Studnicki, Fisher. Acquisition, analysis, or interpretation of the data: all authors. Drafting of manuscript: Studnicki, Fisher. Critical revision of manuscript for important intellectual content: all authors. Statistical analysis: Studnicki, Fisher. Obtained funding: Studnicki. Administrative, technical, or material support: Fisher. Study supervision: Studnicki.
      No financial disclosures were reported by the authors of this paper.

      References

        • Middlesex-London Health Unit
        Identifying priority populations: process, recommendations, and next steps. Ontario, LondonSeptember, 2012
        • Dicenso A.
        • Bayley L.
        • Haynes R.B.
        Accessing pre-appraised evidence: fine tuning the 5S model into a 6S model.
        Evid Based Nurs. 2009; 12: 99-101https://doi.org/10.1136/ebn.12.4.99-b
        • National Quality Forum (Commissioned Paper)
        An environmental scan of integrated approaches for defining and measuring total population health by the clinical system, the government public health system and stakeholder organizations. Authors.
        Jacobson DM and Teutsch S. June 2012;
        • Shah S.N.
        • Russo E.T.
        • Earl T.R.
        • Kuo T.
        Measuring and monitoring progress toward health equity: local challenges for public health.
        Prev Chronic Dis. 2014; 11: 130440https://doi.org/10.5888/pcd11.130440
        • Wolfson M.C.
        Notes on measurement and accountability.
        Presentation to the Institute of Medicine Committee on Public Health Strategies to Improve Health. Meeting 2. January 2010;
        • Studnicki J.
        • Fisher J.W.
        Determining community health status priorities in an online analytic (OLAP) environment.
        Online J Public Health Inform. 2013; 2