Advertisement

Association Between Systolic Blood Pressure and Cardiovascular Inpatient Cost Moderated by Peer-Support Intervention Among Adult Patients With Type 2 Diabetes: A 2-Cohort Study

  • Dahai Yu
    Affiliations
    Department of Nephrology, The First Affiliated Hospital, Zhengzhou University, Zhengzhou, China

    Arthritis Research UK Primary Care Centre, Research Institute for Primary Care & Health Sciences, Keele University, Keele, United Kingdom
    Search for articles by this author
  • Yamei Cai
    Affiliations
    Department of Nephrology, The First Affiliated Hospital, Zhengzhou University, Zhengzhou, China
    Search for articles by this author
  • Jonathan Graffy
    Affiliations
    Primary Care Unit, Department of Public Health and Primary Care, University of Cambridge, Cambridge, Cambridgeshire, United Kingdom
    Search for articles by this author
  • Daniel Holman
    Affiliations
    Department of Sociological Studies, University of Sheffield, Sheffield, United Kingdom
    Search for articles by this author
  • Zhanzheng Zhao
    Correspondence
    Addresses for correspondence: Zhanzheng Zhao MD, Department of Nephrology, The First Affiliated Hospital, Zhengzhou University, Zhengzhou, 450052, China. David Simmons MD, Macarthur Clinical School, School of Medicine, Western Sydney University, Locked Bag 1797, Campbelltown, New South Wales 2751, Australia.
    Affiliations
    Department of Nephrology, The First Affiliated Hospital, Zhengzhou University, Zhengzhou, China
    Search for articles by this author
  • David Simmons
    Correspondence
    Addresses for correspondence: Zhanzheng Zhao MD, Department of Nephrology, The First Affiliated Hospital, Zhengzhou University, Zhengzhou, 450052, China. David Simmons MD, Macarthur Clinical School, School of Medicine, Western Sydney University, Locked Bag 1797, Campbelltown, New South Wales 2751, Australia.
    Affiliations
    Department of Nephrology, The First Affiliated Hospital, Zhengzhou University, Zhengzhou, China

    Western Sydney University, Campbelltown, Sydney New South Wales, Australia
    Search for articles by this author
Open AccessPublished:August 10, 2020DOI:https://doi.org/10.1016/j.jcjd.2020.07.008

      Abstract

      Objectives

      People with type 2 diabetes and increased systolic blood pressure (SBP) are at high risk of cardiovascular disease (CVD). In this study, we aimed to investigate the association between CVD-related hospital payments and SBP and tested whether this association is influenced by diabetes peer support.

      Methods

      Two cohorts comprising people with type 2 diabetes were included in the study. The first cohort comprised 4,704 patients with type 2 diabetes assessed between 2008 and 2009 from 18 general practices in Cambridgeshire and followed up to 2009–2011. The second cohort comprised 1,121 patients with type 2 diabetes from post-trial follow-up data, recruited between 2011 and 2012 and followed up to 2015. SBP was measured at baseline. Inpatient payments for CVD hospitalization within 2 years since baseline was the main outcome. The impact of 1:1, group or combined diabetes peer support and usual care were investigated in the second cohort. Adjusted mean CVD inpatient payments per person were estimated using a 2-part model after adjusting for baseline characteristics.

      Results

      A “hockey-stick” relationship between baseline SBP and estimated CVD inpatient payment was identified in both cohorts, with a threshold at 133 to 141 mmHg, suggesting increased payments for patients with SBP below and above the threshold. The combined peer-support intervention altered the aforementioned association, with no increased payment with SBP above the threshold, and payment slightly decreased with SBP beyond the threshold.

      Conclusions

      SBP maintained between 133 and 141 mmHg is associated with the lowest CVD disease management costs for patients with type 2 diabetes. Combined peer-support intervention could significantly decrease CVD-related hospital payments.

      Résumé

      Objectifs

      Les personnes ayant le diabète de type 2 et une augmentation de la pression artérielle systolique (PAS) sont exposées à un risque élevé de maladies cardiovasculaires (MCV). Dans la présente étude, nous avions pour objectif d’examiner l’association entre les paiements des frais d’hospitalisation liés à la MCV et la PAS, et de vérifier si l’entraide entre pairs diabétiques influence cette association.

      Méthodes

      Nous avons inclus à l’étude 2 cohortes de personnes atteintes du diabète de type 2. La première cohorte regroupait 4704 patients atteints du diabète de type 2 de 18 cabinets de généralistes de Cambridgeshire qui avaient été évalués entre 2008 et 2009 et suivis jusqu’en 2009–2011. La deuxième cohorte regroupait 1121 patients atteints du diabète de type 2 qui avaient été recrutés entre 2011 et 2012 et suivis jusqu’en 2015 et dont les renseignements provenaient des données du suivi post-étude. Nous avons mesuré la PAS au début. Les paiements des frais d’hospitalisation liés à la MCV dans les 2 ans depuis le début étaient le critère principal d’évaluation. Nous avons examiné les répercussions de l’entraide entre pairs diabétiques 1:1, combinée ou en groupe et les soins courants de la seconde cohorte. Nous avons estimé à l’aide d’un modèle en 2 parties les paiements moyens ajustés des frais d’hospitalisation liés à la MCV par personne après l’ajustement des caractéristiques initiales.

      Résultats

      Nous avons établi une relation en « bâton de hockey » entre la PAS initiale et le paiement estimé des frais d’hospitalisation liés à la MCV dans les 2 cohortes, selon un seuil de 133 à 141 mmHg, c’est-à-dire à une hausse des paiements chez les patients ayant une PAS sous et au-dessus du seuil. L’intervention combinée de l’entraide entre pairs a modifié l’association mentionnée plus haut, soit aucune augmentation de paiement lors de PAS au-dessus du seuil et une légère diminution de paiement lors de PAS au-delà du seuil.

      Conclusions

      La PAS maintenue entre 133 et 141 mmHg est associée aux coûts de prise en charge des MCV les plus faibles chez les patients atteints de diabète de type 2. L’intervention combinée de l’entraide entre pairs pourrait diminuer de façon significative les paiements des frais d’hospitalisation liés à la MCV.

      Keywords

      Mots clés

      • It is unclear whether systolic blood pressure level impacts cardiovascular disease inpatient costs and could be altered by peer support in people with diabetes.
      • The association between systolic blood pressure and cardiovascular disease inpatient payment showed a “hockey-stick” shape with a threshold at 133 to 141 mmHg.
      • A novel 2-part model revealed that combined peer-support intervention altered this association.

      Introduction

      The rapid increase in prevalence and health costs associated with type 2 diabetes has been observed worldwide (
      American Diabetes Association
      Executive summary: Standards of medical care in diabetes---2012.
      ). It has been estimated that the risk of hospitalization is 2-fold higher in people with diabetes vs those without diabetes, and the proportion with diabetes is >10% in those admitted to hospital at any one time (
      • Sampson M.J.
      • Dozio N.
      • Ferguson B.
      • Dhatariya K.
      Total and excess bed occupancy by age, specialty and insulin use for nearly one million diabetes patients discharged from all English acute hospitals.
      ). Among some specific age strata, this proportion is >20% (
      • Simmons D.
      • English P.
      • Robins P.
      • Craig A.
      • Addicott R.
      Should diabetes be commissioned through multidisciplinary networks, rather than practice based commissioning?.
      ). The associated costs of excess admissions, as well as increased costs per admission, are significant contributors to the financial burden borne by health-care systems from diabetes and often reflect preventable morbidity suffered by patients (
      • Vamos E.P.
      • Millett C.
      • Parsons C.
      • Aylin P.
      • Majeed A.
      • Bottle A.
      Nationwide study on trends in hospital admissions for major cardiovascular events and procedures among people with and without diabetes in England, 2004–2009.
      ). Systolic blood pressure (SBP), as the most common modifiable risk factor, has been found to be associated both with cardiovascular disease (CVD) mortality and CVD hospitalization among people with type 2 diabetes (
      • Yamout H.
      • Bakris G.L.
      In search for the 'sweet spot' for blood pressure level in diabetes.
      ,
      • Yu D.
      • Simmons D.
      Association between blood pressure and risk of cardiovascular hospital admissions among people with type 2 diabetes.
      ). However, no established association between SBP and inpatient cost due to CVD hospitalization has been shown among people with type 2 diabetes. Although a target SBP has been agreed upon to lower the risk of, for example, CVD mortality and CVD hospitalization, it is unclear whether this threshold has an impact on inpatient costs due to CVD mortality.
      Diabetes peer support involves people with diabetes assisting each other to improve their social, mental and physical well-being. Peer support can be provided through individual or group approaches and by either face-to-face or telephone or online contact. It is generally seen as a low-cost intervention and has been suggested to reduce health-care costs (
      • Dale J.R.
      • Williams S.M.
      • Bowyer V.
      What is the effect of peer support on diabetes outcomes in adults? A systematic review.
      ). Some studies have reported that peer support can reduce health-care costs among people with type 2 diabetes (
      • Johansson T.
      • Keller S.
      • Sonnichsen A.C.
      • Weitgasser R.
      Cost analysis of a peer support programme for patients with type 2 diabetes: A secondary analysis of a controlled trial.
      ,
      • Burton J.
      • Eggleston B.
      • Brenner J.
      • Truchil A.
      • Zulkiewicz B.A.
      • Lewis M.A.
      Community-based health education programs designed to improve clinical measures are unlikely to reduce short-term costs or utilization without additional features targeting these outcomes.
      ). The RAndomised controlled trial of Peer Support In type 2 Diabetes (RAPSID) was the largest randomized, controlled trial (RCT) of type 2 diabetes peer support to date (
      • Simmons D.
      • Prevost A.T.
      • Bunn C.
      • et al.
      Impact of community based peer support in type 2 diabetes: A cluster randomised controlled trial of individual and/or group approaches.
      ). The intervention was recently shown to be cost-effective during the trial based on self-reported costs (
      • Simmons D.
      • Cohn S.
      • Bunn C.
      • et al.
      Testing a peer support intervention for people with type 2 diabetes: A pilot for a randomised controlled trial.
      ) and also from prospective hospital costs (
      • Yu D.
      • Cai Y.
      • Graffy J.
      • Holman D.
      • Zhao Z.
      • Simmons D.
      Development and external validation of risk scores for cardiovascular hospitalization and rehospitalization in patients with diabetes.
      ). In RAPSID, group peer support was associated with 2- to 3-mmHg lower SBP; however, it was unclear whether this was a mediator of the reduction in inpatient costs, and whether this was through an effect on CVD hospitalization specifically.
      In this study, we investigated the association between SBP measured in primary care settings and inpatient cost for CVD hospitalization over the next 2 years, accounting for the risk of hospitalization among people with type 2 diabetes, using data from 2 cohorts. We then examined the impact of peer support on this association.

      Methods

      Data source and study population

      We followed the methods of Yu et al for data and data collection (
      • Yu D.
      • Cai Y.
      • Graffy J.
      • Holman D.
      • Zhao Z.
      • Simmons D.
      Development and external validation of risk scores for cardiovascular hospitalization and rehospitalization in patients with diabetes.
      ). Briefly, we utilized 2 cohorts from Cambridgeshire, England: one (main cohort) based on the electronic health record data from primary care settings to develop our CVD hospitalization and rehospitalization risk scores, and another (replication cohort) based on post-trial cohort data for external validation.

      Main cohort

      Patient lists from 18 general practices across Cambridgeshire, England, in 2008–2009 were collated and linked with hospital admissions (Secondary Uses Service) data as part of an evaluation of diabetes care across the county by the local health board, the National Health Service (NHS) Cambridgeshire. This cohort was limited to volunteer practices using the Egton Medical Information Systems general practitioner software system, from which a predefined set of data could be extracted. There was no systematic selection process for these surgeries, and data extracted were for their entire diabetes population. All patients with diabetes had follow-up hospitalization data to 2010–2011. Hospital admissions to the NHS and private hospitals within and outside Cambridgeshire were followed up. No personal identifiers were released to researchers, and all subsequent analyses were conducted on anonymized data sets. Baseline blood pressure and clinical measurements were recorded as part of clinical practice in primary care settings (
      • Yu D.
      • Simmons D.
      Association between blood pressure and risk of cardiovascular hospital admissions among people with type 2 diabetes.
      ).

      Replication cohort

      The design and methods of the RAPSID have been published previously (
      • Simmons D.
      • Prevost A.T.
      • Bunn C.
      • et al.
      Impact of community based peer support in type 2 diabetes: A cluster randomised controlled trial of individual and/or group approaches.
      ), including its Consolidated Standards of Reporting Trials diagram and the results of its primary outcomes (
      • Simmons D.
      • Prevost A.T.
      • Bunn C.
      • et al.
      Impact of community based peer support in type 2 diabetes: A cluster randomised controlled trial of individual and/or group approaches.
      ). Briefly, RAPSID was a 2×2 factorial cluster RCT comparing 4 groups: controls, 1:1 (individual) peer support, group peer support and combined 1:1 and group peer support among patients with type 2 diabetes. Participants had their diabetes for at least 12 months and those with dementia or mental illness were excluded. Participants were recruited from communities across Cambridgeshire and neighbouring areas of Essex and Hertfordshire. Follow-up data were only available for participants in Cambridgeshire and neighbouring areas of Hertfordshire that are served by the Cambridgeshire and Peterborough Clinical Commissioning Group (CCG). Clusters were defined by local government (parish council) boundaries. The intervention was developed following a pilot, using a framework defined by Peers for Progress (
      • Simmons D.
      • Cohn S.
      • Bunn C.
      • et al.
      Testing a peer support intervention for people with type 2 diabetes: A pilot for a randomised controlled trial.
      ). Peers facilitating peer support were termed peer-support facilitators and their selection, training, support and the overall program have been described elsewhere (
      • Simmons D.
      • Prevost A.T.
      • Bunn C.
      • et al.
      Impact of community based peer support in type 2 diabetes: A cluster randomised controlled trial of individual and/or group approaches.
      ). The intervention lasted 8 to 12 months and was commenced and concluded, cluster by cluster, between June 2, 2011 to April 12, 2012.
      Demographics, blood pressure, glycated hemoglobin (A1C) and lipid profile information were collected at baseline. Blood pressure was measured using an Omron electronic blood pressure monitor (Model 705IT, Omron Corp, Kyoto, Japan) (). Each participant was followed up until June 2015 (0.91 to 4.07 years of follow up from beginning/entry into the trial). Hospitalization (NHS hospitals and private hospitals), accident and emergency and outpatient visits within/outside Cambridgeshire and the included areas of Hertfordshire were collected through the Cambridgeshire and Peterborough CCG (
      • Simmons D.
      • Yu D.
      • Wenzel H.
      Changes in hospital admissions and inpatient tariff associated with a diabetes integrated care initiative: Preliminary findings.
      ) as well as elective/nonelective status and International Classification of Diseases10th revision (ICD-10) codes (
      • Yu D.
      • Simmons D.
      Association between blood pressure and risk of cardiovascular hospital admissions among people with type 2 diabetes.
      ).

      Ethics approval

      Ethics approval was received from the Cambridgeshire REC2 Committee (10/H0308/72), and signed consent included agreement for access to hospital data.

      Defining CVD hospitalization

      The primary outcome of the study was having at least 1 hospitalization with CVD as the primary diagnosis (ICD-10 codes I20 to I25, I60 to I69 and I73 in the first ICD field) over the 2-year follow-up period.

      Statistical analysis

      A large proportion of the diabetes population do not attend hospital as an inpatient or outpatient in any given year and, therefore, health-care payment data demonstrate a skewed utilization/payment pattern (
      • Andrade L.F.
      • Rapp T.
      • Sevilla-Dedieu C.
      Exploring the determinants of endocrinologist visits by patients with diabetes.
      ). To take into account the problem of “zero mass” and skewed outcomes, the demand functions were modelled using a 2-part model (
      • Egede L.E.
      • Walker R.J.
      • Bishu K.
      • Dismuke C.E.
      Trends in costs of depression in adults with diabetes in the United States: Medical Expenditure Panel Survey, 2004–2011.
      ). In this 2-part model, a probit model was estimated for the probability of observing “zero” vs positive medical expenditure. Positive medical expenditure is defined as any health-care expenditure greater than zero. A generalized linear model was estimated, conditional on having health-care expenditure. The generalized linear model was used instead of log ordinary least-squares regression, because it relaxes the normality and homoscedasticity assumption and avoids bias associated with retransforming to the raw scale (
      • Bruno G.
      • Picariello R.
      • Petrelli A.
      • et al.
      Direct costs in diabetic and non diabetic people: The population-based Turin study.
      ). The results of the modified Park test verified that the use of a gamma distribution, with a log link, was the best-fitted generalized linear model for consistent estimation of coefficients (
      • Smith V.A.
      • Preisser J.S.
      • Neelon B.
      • Maciejewski M.L.
      A marginalized two-part model for semicontinuous data.
      ). The variance inflation factor for all predictors used in the 2-part model indicated no presence of multicollinearity (
      • Rudisill C.
      • Charlton J.
      • Booth H.P.
      • Gulliford M.C.
      Are healthcare costs from obesity associated with body mass index, comorbidity or depression? Cohort study using electronic health records.
      ). The F-test result for the 2-part regression models was found to be significant, which indicated the overall significance of the regression model. Predicted inpatient cost was estimated in the 2-part model by the level of baseline SBP with adjustment of other covariables. The 95% confidence intervals for estimated payments were estimated by a bootstrap process with 1,000 samples. Analysis restricted analyses in each financial year were carried out as sensitivity analyses. All analyses were performed with STATA version 14.0 (StataCorp, College Station, Texas, United States).

      Results

      In our main cohort, we analyzed information on 4,704 type 2 diabetes patients with 588 CVD hospitalizations within 2 years. Our replication cohort had information on 1,121 type 2 diabetes patients with 183 CVD hospitalizations. Table 1 summarizes the basic characteristics and clinical measurements of the study population. Patients with type 2 diabetes in both cohorts had similar age, gender, blood pressure and total cholesterol. Patients in the main cohort had a higher level of high-density lipoprotein, low-density lipoprotein and A1C. Compared with the main cohort, those in the replication cohort were more likely to be prescribed lipid-lowering medications. Baseline data for the 4 groups of the replication cohort were well matched (Table 1). The sample size of the main cohort and intervention groups of the replication cohort, characteristics of participants and median cost in 2 cohorts by baseline SBP categories (<120, 120 to 129, 130 to 139, 140 to 149, 150 to 159 and >160 mmHg) are presented in Table 2.
      Table 1Baseline characteristics of study cohorts
      Main cohortReplication cohort
      AllControl1:1GroupCombined
      N4,7021,121291261288281
      CVD hospitalization, n (%)588 (12.5)183 (16.3)59 (20.3)45 (17.2)42 (14.6)37 (13.2)
      Age, years65.0±16.365.5±11.465.9±12.865.3±9.865.8±11.965.0±10.4
      Female, n (%)1,919 (40.8)444 (39.6)122 (41.9)109 (41.8)101 (35.1)112 (39.9)
      Systolic blood pressure, mmHg134.5±16.0139.7±20.2140.0±20.6140.4±20.6140.8±19.5137.9±20.3
      Diastolic blood pressure, mmHg76.3±10.075.5±11.575.0±11.675.8±10.975.1±11.375.6±11.9
      Total cholesterol, mmol/L4.3±1.24.2±1.74.3±1.54.3±1.34.1±2.04.3±1.7
      High-density lipoprotein, mmol/L1.3±0.61.1±1.21.2±0.91.2±1.01.0±1.51.1±1.1
      Low-density lipoprotein, mmol/L2.5±1.41.4±3.01.3±3.21.5±2.81.5±2.81.5±3.0
      Body mass index, kg/m230.8±6.932.2±6.032.3±6.032.6±6.532.0±5.932.2±5.9
      A1C, mmol/mol61.5±17.256.2±15.155.6±16.256.5±15.057.3±14.755.3±13.8
      Lipid-lowering treatment, n (%)3,342 (71.4)731 (65.2)180 (61.9)173 (66.3)191 (66.3)187 (66.6)
      A1C, glycated hemoglobin; CVD, cardiovascular disease.
      Note: Data expressed as number, number (%) or mean ± standard deviation.
      Table 2Distribution of baseline characteristics and inpatient cost due to CVD hospitalization in the study cohorts
      Category of systolic blood pressure, mmHg
      <120120–129130–139140–149150–159≥160
      Main cohort5647951,2041,059324756
      Replication cohort, overall134 (12.0)174 (15.5)244 (21.8)255 (22.8)167 (14.9)147 (13.1)
      Replication cohort, control35 (11.9)46 (15.8)64 (22.1)68 (23.5)46 (15.8)32 (10.9)
      Replication cohort, group33 (12.5)35 (13.6)50 (19.1)66 (25.4)35 (13.6)41 (15.8)
      Replication cohort, 1:130 (10.5)43 (15.0)67 (23.3)62 (21.6)43 (15.0)42 (14.6)
      Replication cohort, combined37 (13.0)50 (17.7)63 (22.4)58 (20.6)43 (15.2)31 (11.2)
      Age, years59.9±18.062.7±15.865.1±14.067.7±12.968.6±13.265.7±19.2
      Female, n (%)244 (35.0)382 (39.4)611 (42.2)572 (43.5)208 (42.3)347 (38.4)
      Systolic blood pressure, mmHg110.3±8.3123.8±3.1133.7±3.1143.0±3.0153.4±3.1169.4±10.9
      Diastolic blood pressure, mmHg68.0±9.273.4±8.476.0±8.378.5±8.781.2±9.585.0±11.3
      Total cholesterol, mmol/L4.2±1.24.2±1.14.3±1.14.3±1.24.5±1.24.6±1.3
      High-density lipoprotein, mmol/L1.3±0.51.2±0.41.3±0.51.3±0.51.3±0.51.3±0.5
      Low-density lipoprotein, mmol/L2.4±1.02.4±1.02.5±1.02.5±1.02.5±1.02.6±1.0
      Body mass index, kg/m229.8±6.330.5±6.731.5±6.931.5±6.631.7±6.430.4±7.3
      A1C, mmol/mol61.5±19.560.2±16.960.9±15.660.5±16.061.0±15.862.4±16.7
      Lipid-lowering treatment, n (%)459 (65.8)688 (71.0)1,066 (73.6)1,010 (76.9)364 (74.2)484 (53.6)
      People with zero payment, n (%)602 (11.9)865 (29.0)1,263 (25.0)1,109 (22.0)382 (7.6)832 (16.5)
      Cost among people with nonzero payment, £2,436.37 (629.40–5,277.45)2,017.75 (763.53–3,561.62)1,781.41 (644.42–4,931.43)2,507.73 (1,318.58–4,786.15)2,801.81 (893.70–4,008.91)3,485.46 (1,362.57–4,956.08)
      A1C, glycated hemoglobin; CVD, cardiovascular disease.
      Note: Data expressed as number, number (%), mean ± standard deviation or median (interquartile range).
      As shown in Supplementary Table 1, inpatient cost data from CVD hospitalization were typically skewed due to the mass of “zero” payments and a relatively small proportion of patients incurring extremely high expenditure. Of the participants in the main and replication cohort, 87.5% and 83.7%, respectively, were not hospitalized due to CVD diseases over the 2-year of follow up. Within the replication cohort, 79.7% of controls and 82.8%, 85.4% and 86.8% of patients in the 1:1 group and combined intervention groups, respectively, were not hospitalized due to CVD disease. Among patients hospitalized due to CVD diseases, median inpatient costs were £4348.35 (interquartile range [IQR]: 1,623.50 to 8,766.75) and £2,430.72 (IQR: 793.06 to 4,026.20) for the main and replication cohorts, respectively. With the replication cohort, median inpatient costs were £2,419.60 (IQR: 1,006.91 to 4,387.66), £2,489.40 (IQR: 770.69 to 4,387.66), £1,963.56 (IQR: 714.93 to 4,032.55) and £2,436.00 (IQR: 885.19 to 3,473.12) for the controls, 1:1 group and combined intervention groups, respectively. Compared with patients without inpatient costs due to CVD hospitalization, patients with such costs were more likely to be older; male; have higher SBP, body mass index and A1C; and less likely to have been prescribed lipid-lowering treatment (Supplementary Table 2).

      Results from 2-part model

      Dose–response relationship curves between SBP and predicted inpatient cost for CVD hospitalization derived from the 2-part models after accounting for the risk of CVD hospitalization with adjustment of covariables in Table 1 are presented in Figure 1 for the main cohort and the replication cohort. SBP was nonlinearly associated with adjusted predicted inpatient cost for CVD hospitalization (linearity test: p<0.00001 for all) in both the main and replication cohorts. The threshold was estimated at 137 (133 to 141) mmHg for SBP in both the main sample and replication sample, with consistent stable adjusted predicted inpatient cost for CVD hospitalization below the threshold and increased predicted inpatient cost above the threshold.
      Figure thumbnail gr1
      Figure 1Adjusted association between baseline systolic blood pressure and predicted inpatient cost due to CV disease hospitalization in the main cohort and replication cohort. CV, cardiovascular.
      Within the replication cohort, dose–response relationship curves between SBP and adjusted predicted inpatient cost for CVD hospitalization in each group are presented in Figure 2. A nonlinear association between SBP and adjusted predicted inpatient cost was found in the control, 1:1 and intervention groups (linearity test: p<0.00001 for all). The threshold at 137 (133 to 141) mmHg for SBP was found consistently in each group, with consistent stable adjusted predicted inpatient cost for CVD hospitalization below the threshold and increased predicted inpatient cost above the threshold. In the combined intervention group, the adjusted predicted inpatient cost was linearly stable as SBP increased (linearity test: p=0.05263). Associations between baseline systolic blood pressure and predicted inpatient cost due to CVD hospitalization in those receiving 1:1 peer support and among the others within the replication cohort are presented in Supplementary Figure 1. For each baseline SBP level, patients who received the 1:1 peer-support intervention (as 1:1 alone or as the combined intervention) were more likely to have lower inpatient costs due to CVD hospitalization.
      Figure thumbnail gr2
      Figure 2Adjusted association between baseline systolic blood pressure and predicted inpatient cost due to CV disease hospitalization in groups of the replication cohort. CV, cardiovascular.

      Discussion

      In this study, using 2 prospective cohorts, we found a nonlinear association between SBP measured in United Kingdom populations with type 2 diabetes and the adjusted predicted inpatient cost for CVD hospitalization over 2 years of follow up, after accounting for the risk of CVD hospitalization both in the main and replication cohort. Further investigation revealed that SBP <137 mmHg was associated with stable lowest inpatient cost, and inpatient cost increased with an increase in SBP >137 mmHg. The peer-support intervention, especially group intervention combined with 1:1 support, had a significant impact on the association between inpatient cost for CVD hospitalization and SBP.

      Comparison with previous studies

      It is well established that SBP is the major determinant of CVD risk in the population >50 years of age (
      • Bangalore S.
      • Messerli F.H.
      • Wun C.C.
      • et al.
      J-curve revisited: An analysis of blood pressure and cardiovascular events in the Treating to New Targets (TNT) trial.
      ). In patients with type 2 diabetes, previous studies have revealed a J-shape relationship between SBP and CVD event risk; for example, the United Kingdom Prospective Diabetes Study (
      • Bloch M.J.
      • Basile J.N.
      Analysis of recent papers in hypertension. Lack of legacy effect with more intensive blood pressure control in the long-term follow-up of the United Kingdom Prospective Diabetes Study.
      ) showed a lowered CVD event rate with an attained lower blood pressure goal of 144/82 mm Hg. The International Verapamil SR—Trandolapril (
      • Bangalore S.
      • Gong Y.
      • Cooper-DeHoff R.M.
      • Pepine C.J.
      • Messerli F.H.
      2014 Eighth Joint National Committee panel recommendation for blood pressure targets revisited: Results from the INVEST study.
      ) and the Avoiding CVD Events in Combination Therapy in Patients Living with Systolic Hypertension (
      • Bakris G.
      • Hester A.
      • Weber M.
      • et al.
      The diabetes subgroup baseline characteristics of the Avoiding Cardiovascular Events Through Combination Therapy in Patients Living With Systolic Hypertension (ACCOMPLISH) trial.
      ) trials also failed to demonstrate a CVD outcome benefit at a blood pressure of <130/80 mmHg. We have previously shown that an SBP between 133 and 141 mmHg was associated with the lowest risk of CVD hospitalization among patients with type 2 diabetes (
      • Yu D.
      • Simmons D.
      Association between blood pressure and risk of cardiovascular hospital admissions among people with type 2 diabetes.
      ). However, it was not clear whether this J-shape relationship exists between SBP and inpatient costs for CVD hospitalization as most studies analyzed health cost/payments that had a skewed distribution. Ours is the first study among patients with type 2 diabetes mellitus, following adjustment for the individual probability of being hospitalized, and we have now shown a “hockey-stick”-shaped relationship between SBP and CVD inpatient payment. This finding suggests that CVD inpatient payments are stable for SBP below 133 to 141 mmHg and linearly increase above this range. This in turn supports an SBP target between 133 and 141 mmHg to minimize future risk of CVD hospitalization and associated inpatient payments.
      Although we have shown that CVD hospital payments increase with a baseline SBP above 133 to 144 mmHg, this was not observed in the 2-year posttrial period for RAPSID intervention participants. In RAPSID, peer group support was associated with a significant reduction in SBP after 8- to 12-month follow up from baseline and we speculate that it was this lower SBP that was responsible for this finding. Hospitalization was shown to be reduced in Hong Kong with peer support among those who had high diabetes distress (
      • Chan J.C.
      • Sui Y.
      • Oldenburg B.
      • et al.
      Effects of telephone-based peer support in patients with type 2 diabetes mellitus receiving integrated care: A randomized clinical trial.
      ). We have not been able to elucidate the mechanism behind the lower SBP in RAPSID and have excluded a greater effect among those with high diabetes distress and medication adherence. There was also no evidence of changes in lifestyle as measured by questionnaires, or crudely by body weight (a small reduction in waist circumference was found in the per-protocol analyses). The current finding of reduced CVD hospitalization costs does provide some validity that the lower SBP described was not simply due to chance.
      The data suggest that the peer-support intervention was associated with a reduced inpatient payment; however, in the 2-year post-trial follow up among patients in the combined intervention group, CVD inpatient payment did not increase along with the increase of SBP, especially beyond 133 to 141 mmHg. The slight reduction in CVD inpatient payment suggests that patients whose SBP was beyond 133 to 141 mmHg were less likely to trigger the CVD hospitalization, with this primarily due to the combined peer-support intervention. It could be that patients in the combined intervention group adhered to a healthy lifestyle in the post-trial follow up, which could have impacted patients’ obesity status and then SBP, as observed in the trial follow up. However, no post-trial obesity measurement was done to validate this hypothesis. In the trial follow up the SBP reduction could not be explained by increased medication adherence, as this was previously found to be unchanged (
      • Simmons D.
      • Prevost A.T.
      • Bunn C.
      • et al.
      Impact of community based peer support in type 2 diabetes: A cluster randomised controlled trial of individual and/or group approaches.
      ). It is unclear whether the antihypertensive treatment adherence pattern was modified in the post-trial follow-up restricted by the post-trial information on the medication adherence.

      Strength and limitations

      One strength of our analysis is that the association between SBP and CVD inpatient payment was examined in 2 independent cohorts. Another strength is the minimal information bias, with the outcome used, recorded inpatient payments, having been fully recorded by the CCG (
      • Simmons D.
      • Yu D.
      • Wenzel H.
      Changes in hospital admissions and inpatient tariff associated with a diabetes integrated care initiative: Preliminary findings.
      ). In particular, as these are payment details, both NHS hospitals and private hospital admissions were able to be included. There would have been some loss for patients where no component of care was paid for by the CCG.
      Some limitations must also be considered in the interpretation of our findings. Unlike pharmaceutical interventions, where adherence can be assessed using pill counters, it is difficult to evaluate the magnitude of peer-support intervention on an individual level, and, although we did record attendance and telephone calls, we did not assess engagement. The payment/savings from similar peer-support interventions should be further investigated in other post-trial observation studies. Another limitation is the inconsistent blood pressure measurement approaches between those used in primary care (main cohort) and those used in the trial (replication cohort) in terms of attended vs unattended, standardized protocol vs usual measurement and automated vs mercury sphygmomanometer. A final limitation of our study is that we obtained information on the activities of the participants after the trial was completed. All participants were sent the results, and we are aware that some intervention (e.g. peer-support groups) continued, including support from the Diabetes UK Type 2 Together program (
      • Simmons D.
      • Cohn S.
      • Bunn C.
      • et al.
      Testing a peer support intervention for people with type 2 diabetes: A pilot for a randomised controlled trial.
      ).
      In conclusion, to our knowledge, this study is the first to examine the prospective association between SBP and 2-year estimated CVD inpatient payments. A “hockey-stick” relationship between SBP and 2-year estimated CVD inpatient payment was identified in 2 independent cohorts, with a consistent threshold at 133 to 141 mmHg and a linearly increasing payment beyond the threshold. An alteration in this relationship after a combined peer-support intervention (group and 1:1 interventions) is suggested by the lack of increase in estimated CVD payment. Our findings suggest that, among people with type 2 diabetes, blood pressure management should target an SBP of 133 to 141 mmHg. Integration of this threshold into clinical practice could lower both individual risk of and associated payments for CVD hospitalization.

      Acknowledgments

      The authors thank Toby Prevost, Chris Bunn, Simon Cohn, Sarah Donald, Charlotte Paddison, Candice Ward, Peers for Progress, West Anglia CLRN, Cambridgeshire & Peterborough PCT, Primary Care Research Network–East of England, Eastern Diabetes Research Network, MRC Epidemiology Unit, participating general practices, Jackie Williams, Caroline Taylor, Kym Mercer, Kevin Baker, Ben Bowers, Kalsoom Akhter (CUH Wolfson Diabetes & Endocrinology Clinic), James Brimicombe (Cambridge University), Kim Birch of Trumpington St General Practice, CUH Wolfson Diabetes & Endocrinology Clinic Educators, the RAPSID Patient Committee (Phillip Jones, Liz Carvlin, Roger Smith) and the peers and peer-support participants. The main cohort derived from an independent research funded by the National Institute for Health Research (NIHR) under its Research for Patient Benefit (RfPB) Programme (Grant Ref. No. PB-PG-0808-17303). The RAPSID was supported by Peers for Progress (peersforprogress.org) and the NIHR for Patient Benefit Programme (Ref. No. PB-PG-0610-22311). The funders had no role in the study design, data collection and analysis, decision to publish or preparation of the manuscript. Recruitment and follow-up assessments were largely undertaken by the NIHR Primary Care Research Network and additional funding to support recruitment was provided by the NIHR West Anglia Comprehensive Local Research Network. NHS Cambridgeshire funded research nurses to link with the peer-support facilitators. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.

      Author Disclosures

      Conflicts of interest: None.

      Author Contributions

      D.Y., Y.C., Z.Z. and D.S. contributed to the conception and design of this work and to interpretation of the data. D.Y. performed the data management and analysis and drafted the initial manuscript. D.Y., Y.C., Z.Z., D.H., J.G. and D.S. reviewed, revised and approved the final manuscript.

      Supplementary Material

      Supplementary Table 1Distribution of baseline characteristics and inpatient cost due to CVD hospitalization in the study cohorts
      Median cost (interquartile range) among people with nonzero payment, £People with zero payment, N (%)
      Main cohort4,348.35 (1,623.50‒8,766.75)4,116 (87.5%)
      Replication cohort, overall2,430.72 (793.06‒4,026.20)938 (83.7%)
      Replication cohort, control2,419.60 (1,006.91‒4,387.66)232 (79.7%)
      Replication cohort, group2,489.40 (770.69‒4,387.66)216 (82.8%)
      Replication cohort, 1:11,963.56 (714.93‒4,032.55)246 (85.4%%)
      Replication cohort, combined2,436.00 (885.19‒3,473.12)244 (86.8%)
      CVD, cardiovascular disease.
      Supplementary Table 2Distribution of baseline characteristics among patients with and without inpatient cost due to CVD hospitalization
      People with paymentPeople with zero paymentP value
      N7695,054
      Age, years71.4±11.864.0±16.0<0.0001
      Female, n (%)282 (36.6)2,084 (41.2)<0.0001
      Systolic blood pressure, mmHg138.3±16.9135.2±16.2<0.0001
      Diastolic blood pressure, mmHg75.3±10.076.4±9.9<0.0001
      Total cholesterol, mmol/L4.3±1.14.3±1.20.0356
      High-density lipoprotein, mmol/L1.2±0.41.3±0.5<0.0001
      Low-density lipoprotein, mmol/L2.3±0.92.5±1.0<0.0001
      Body mass index, kg/m231.9±6.930.8±6.8<0.0001
      A1C, mmol/mol61.7±16.760.7±16.6<0.0001
      Lipid-lowering treatment, n (%)141 (18.3)1,596 (31.6)<0.0001
      A1C, glycated hemoglobin; CVD, cardiovascular disease.
      Figure thumbnail fx1
      Supplementary Figure 1Adjusted association between baseline systolic blood pressure and predicted inpatient cost due to cardiovascular disease hospitalization in those receiving 1:1 support intervention (1:1 and combined intervention) and the others (control and group intervention). CV, cardiovascular.

      References

        • American Diabetes Association
        Executive summary: Standards of medical care in diabetes---2012.
        Diabetes Care. 2012; 35: S4-S10
        • Sampson M.J.
        • Dozio N.
        • Ferguson B.
        • Dhatariya K.
        Total and excess bed occupancy by age, specialty and insulin use for nearly one million diabetes patients discharged from all English acute hospitals.
        Diabetes Res Clin Pract. 2007; 77: 92-98
        • Simmons D.
        • English P.
        • Robins P.
        • Craig A.
        • Addicott R.
        Should diabetes be commissioned through multidisciplinary networks, rather than practice based commissioning?.
        Prim Care Diabetes. 2011; 5: 39-44
        • Vamos E.P.
        • Millett C.
        • Parsons C.
        • Aylin P.
        • Majeed A.
        • Bottle A.
        Nationwide study on trends in hospital admissions for major cardiovascular events and procedures among people with and without diabetes in England, 2004–2009.
        Diabetes Care. 2012; 35: 265-272
        • Yamout H.
        • Bakris G.L.
        In search for the 'sweet spot' for blood pressure level in diabetes.
        Heart. 2014; 100: 1404-1405
        • Yu D.
        • Simmons D.
        Association between blood pressure and risk of cardiovascular hospital admissions among people with type 2 diabetes.
        Heart. 2014; 100: 1444-1449
        • Dale J.R.
        • Williams S.M.
        • Bowyer V.
        What is the effect of peer support on diabetes outcomes in adults? A systematic review.
        Diabet Med. 2012; 29: 1361-1377
        • Johansson T.
        • Keller S.
        • Sonnichsen A.C.
        • Weitgasser R.
        Cost analysis of a peer support programme for patients with type 2 diabetes: A secondary analysis of a controlled trial.
        Eur J Public Health. 2017; 27: 256-261
        • Burton J.
        • Eggleston B.
        • Brenner J.
        • Truchil A.
        • Zulkiewicz B.A.
        • Lewis M.A.
        Community-based health education programs designed to improve clinical measures are unlikely to reduce short-term costs or utilization without additional features targeting these outcomes.
        Popul Health Manag. 2017; 20: 93-98
        • Simmons D.
        • Prevost A.T.
        • Bunn C.
        • et al.
        Impact of community based peer support in type 2 diabetes: A cluster randomised controlled trial of individual and/or group approaches.
        PLoS One. 2015; 10e0120277
        • Simmons D.
        • Cohn S.
        • Bunn C.
        • et al.
        Testing a peer support intervention for people with type 2 diabetes: A pilot for a randomised controlled trial.
        BMC Fam Pract. 2013; 14 (5-2296-14-5)
        • Yu D.
        • Cai Y.
        • Graffy J.
        • Holman D.
        • Zhao Z.
        • Simmons D.
        Development and external validation of risk scores for cardiovascular hospitalization and rehospitalization in patients with diabetes.
        J Clin Endocrinol Metab. 2018; 103: 1122-1129
        • National Institute for Health and Care Excellence
        (Accessed September 18, 2020)
        • Simmons D.
        • Yu D.
        • Wenzel H.
        Changes in hospital admissions and inpatient tariff associated with a diabetes integrated care initiative: Preliminary findings.
        J Diabetes. 2014; 6: 81-89
        • Andrade L.F.
        • Rapp T.
        • Sevilla-Dedieu C.
        Exploring the determinants of endocrinologist visits by patients with diabetes.
        Eur J Health Econ. 2016; 17: 1173-1184
        • Egede L.E.
        • Walker R.J.
        • Bishu K.
        • Dismuke C.E.
        Trends in costs of depression in adults with diabetes in the United States: Medical Expenditure Panel Survey, 2004–2011.
        J Gen Intern Med. 2016; 31: 615-622
        • Bruno G.
        • Picariello R.
        • Petrelli A.
        • et al.
        Direct costs in diabetic and non diabetic people: The population-based Turin study.
        Italy. Nutr Metab Cardiovasc Dis. 2012; 22: 684-690
        • Smith V.A.
        • Preisser J.S.
        • Neelon B.
        • Maciejewski M.L.
        A marginalized two-part model for semicontinuous data.
        Stat Med. 2014; 33: 4891-4903
        • Rudisill C.
        • Charlton J.
        • Booth H.P.
        • Gulliford M.C.
        Are healthcare costs from obesity associated with body mass index, comorbidity or depression? Cohort study using electronic health records.
        Clin Obes. 2016; 6: 225-231
        • Bangalore S.
        • Messerli F.H.
        • Wun C.C.
        • et al.
        J-curve revisited: An analysis of blood pressure and cardiovascular events in the Treating to New Targets (TNT) trial.
        Eur Heart J. 2010; 31: 2897-2908
        • Bloch M.J.
        • Basile J.N.
        Analysis of recent papers in hypertension. Lack of legacy effect with more intensive blood pressure control in the long-term follow-up of the United Kingdom Prospective Diabetes Study.
        J Clin Hypertens. 2009; 11: 46-49
        • Bangalore S.
        • Gong Y.
        • Cooper-DeHoff R.M.
        • Pepine C.J.
        • Messerli F.H.
        2014 Eighth Joint National Committee panel recommendation for blood pressure targets revisited: Results from the INVEST study.
        J Am Coll Cardiol. 2014; 64: 784-793
        • Bakris G.
        • Hester A.
        • Weber M.
        • et al.
        The diabetes subgroup baseline characteristics of the Avoiding Cardiovascular Events Through Combination Therapy in Patients Living With Systolic Hypertension (ACCOMPLISH) trial.
        J Cardiometab Syndr. 2008; 3: 229-233
        • Chan J.C.
        • Sui Y.
        • Oldenburg B.
        • et al.
        Effects of telephone-based peer support in patients with type 2 diabetes mellitus receiving integrated care: A randomized clinical trial.
        JAMA Intern Med. 2014; 174: 972-981