Introduction
Materials and methods
Literature and information search
Eligibility of studies
Data items
Risk of bias (quality) assessment
Data analysis
Results
Literature search yield
Types of study designs, populations and settings
Author year; country | Study design | Main aim of the study | Type and source of data | Sample size (N) and population | Follow-up | COC indicators | NOS score | COC measurement period | Causality discussed | Outcomes associated with COC | Main results of interest |
---|---|---|---|---|---|---|---|---|---|---|---|
Bazemore 2018 ; USA [37] | CS | To examine the relationship between physician-level continuity and healthcare expenditures and hospitalizations | US Medicare claims data; Medicare Quality Payment Program | 1,448,952 Medicare beneficiaries | 2011 | UPC; COCI; MMCI; HI | 9/10 | Unclear | No | Healthcare expenditures and hospitalizations | Higher COC was associated with lower total expenditures (β = –0.151; CI: –0.186 –0.116) and hospitalization rates (OR=0.839, CI:0.787-0.893) |
Blozik 2020; Switzerland [44] | RC | To evaluate continuity of care in Swiss cancer patients based on routine data of mandatory health insurance using four established continuity scales | Helsana health insurance group claims data | 23ʹ515 patients with incident use of antineoplastics | 2014-2017 | UPC; MMCI; COCI; SECON; composite | 9/9 | Single year | Partly, reversed causality accounted in design | Costs, death, and hospitali-zation by type of consul-tations: all, only with GP, with specialists, ambulatory | Higher COC was associated with lower costs for consultations with GP (β=-0.035, p<0.01 for COCI; β=-0.092, p<0.001 for SECON) and mixed for ambulatory consultations (β=0.041, p<0.001 for COCI; β=-0.114, p<0.001 for SECON), while higher costs for all consultation (β=0.087, p<0.001 for COCI; β=0.033, p<0.01 for SECON) and mixed results for the costs for consultations with specialists (β=0.106, p<0.001 for COCI; β=-0.072, p<0.001 for SECON). For hospitalizations the results are mixed: higher COC is associated with higher OR of hospitalizations for all consultations (OR=1.77, p<0.001 for COCI; OR=1.32, p<0.01 for SECON), lower OR for consultations with GP only (insignificant), higher OR for consultations with specialists (insignificant) and mixed results for ambulatory consultations (significant for UPC and MMCI only: OR=0.74, p<0.01 identically for both) |
Chen 2011 ; Taiwan [45] | RC | To examine the effects of COC on healthcare utilization and expenses | NHI claims data set | 48,107 DM patients | 2000-2006 | COCI | 9/9 | Each year of follow-up | Partly, acknowledged in limitations and applied in the design | Healthcare utilization, ED visits and healthcare expenses | High COC compared to low COC was associated with less hospitalizations (OR=0.26, CI:0.25-0.27) and ED visits(OR=0.34, CI:0.33-0.36) and lower healthcare expenses (pharmaceutical expenses: β= –0.14, P <.001, total diabetes-related expenses: β= –0.53, P <.001) |
Chen 2019; Taiwan [46] | RC | To examine the effects of COC on the utilization of follow-up services and outcome of breast cancer patients (stages I–III) in the post-treatment phase of care | Taiwan Cancer Registry, National Healthcare Insurance Database (NHIRD) and the National Register of Deaths | 18,031 patients newly diagnosed with breast cancer (Stage I, II, or III) | 4 periods of 2 years in 2002-2007 | COCI | 8/9 | Each year for 4 follow-up periods | Partly, reversed causality accounted in design | Hospital admission, emergency department visit | Higher oncology COCI was associated with a lower likelihood of hospitalization (OR = 0.78, 95% CI: 0.71–0.85) and emergency department use (OR = 0.88, 95% CI: 0.82–0.95). A higher PCP COCI was also associated with a lower likelihood of hospitalization (OR = 0.77, 95% CI: 0.70–0.85) and emergency department use (OR = 0.75, 95% CI: 0.68–0.82). |
Chen 2020a; Taiwan [47] | RC | To examine and provide quantitative evidence of the effects of COC on end-of-life care among patients with ESRD | National Health Insurance Research Database of Taiwan | 29,095 elderly patients with end-stage renal disease | 2005-2013 | COCI | 6/9 | 6-12 months before death | Partly, reversed causality accounted in design | Healthcare expenditures admissions to the Intensive Care Unit (ICU), visits to the Emergency Room | Better COC would result in lower cumulative expenditures: 1% increase in COCI was associated with an 8%, CI: 7–9% and a 6%, CI: 4%–7% reduction in the total health costs in the 6 and 3 months before death. Patients with perfect provider COCI (provider COCI=1) had lower utilization of ICU: OR=0.78, CI: 0.67-0.90. Patients with perfect site COCI had an OR of 0.74 (CI:0.67-0.83) in the utilization of the ER |
Chen 2020b; Taiwan [48] | RC | To examine the relationship between COC and care coordination, simultaneously evaluate the effects of COC and care coordination on healthcare outcomes, and investigate whether these effects vary across study subjects with different levels of comorbidity | Longitudinal Cohort of Diabetes Patients claims data sets constructed by the National Health Research Institute in Taiwan | 57,965 patients aged 18 years or older who were newly diagnosed with diabetes | 2007-2011 | COCI, Care density | 8/9 | Each year of 3 years follow-up | Partly, reversed causality accounted in design and discussed in limitations | Hospital admission for diabetes or cardiovascular/cerebrovascular conditions | Multi-morbid patients with high- or median-COCI were less likely to be hospitalized for diabetes-related conditions than those in the low-continuity group (OR=0.86, CI: 0.77- 0.96) and (OR=0.90, CI: 0.83-0.98), respectively. For care coordination, the patients in the high- or median-care-density groups were less likely to be hospitalized for diabetes-related conditions than the patients in the low-care-density group (OR=0.84, CI: 0.76-0.93, and OR=0.91, CI:0.84-0.99). |
Cho 2015 ; Korea [38] | CS | To analyze the association between continuity of ambulatory care and hospital admission and determine which continuity index has the best explanatory ability for hospital admission | National Health Insurance Sample (NHIS) | 54,458 participants with type 2 DM | 2009 | UPC; COCI; SECON; ICOC | 10/10 | Single year | No | Hospital admission | Low COC was associated with higher probability to be hospitalized (OR=2.44, CI: 2.17-2.75) |
Cho 2016 ; Korea [49] | RC | To assess the effects of competition among healthcare providers with regard to COC, and the association between COC and hospital admissions | Korean universal health insurance program | 9,997 Asthmatic children | 2009-2013 | UPC | 7/9 | Each year of follow-up | No | Hospital admissions | Lower COC (below mean) is associated with increased hospital admission (OR=2.72, CI:2.14-3.46) |
Christakis 2001a ; USA [50] | RC | To search if COC is associated with better quality of care | Administrative claims data; Washington State Medicaid | 252 Medicaid children with DM (<18 years old) | 1997 | COCI | 7/9 | Single year | No | Hospitalization (inpatient DKA) | High and medium COC was associated with reduced risk of hospitalization for DKA (OR=0.14, CI:0.03-0.67 and OR=0.22, CI:0.05-0.87, respectively) |
Christakis 2001b ; USA [51] | RC | To examine the association between COC and ED visits and hospitalization | Claims data of Group Health Cooperative | 46,097 children | 1993-1998 | COCI | 9/9 | Unclear | Yes, addressed in the design and discussed in limitations | ED use and hospitalization | Lower COC was associated with higher risk of ED visits (HR=1.58, CI:1.49-1.66), ED visits for asthma (HR=1.13, CI:0.82-1.60), hospitalization (HR=1.54, CI:1.33-1.75) and asthma-related hospitalization (HR=1.79, CI:1.21-2.56) |
Christakis 1999 ; USA [52] | RC | To assess whether greater COC was associated with lower ED use | Outpatient teaching clinic at Children’s Hospital and Regional Medical Center, Seattle | 785 Medicaid children (≤19 years) | 1993-1997 | COCI | 7/9 | Unclear | Partly, reverse causality was accounted for in the design | ED use | Higher COC was associated with an decreased ED utilization (HR=0.65, CI:0.50-0.80) |
CHU 2012 ; Taiwan [53] | RC | To examine the association between COC and potentially inappropriate medication use and indirectly with healthcare outcomes and expenses | Longitudinal Health Insurance Database | 51,804 patients (≥65 years) | 2004-2009 | COCI | 9/9 | Each year of follow-up | Partly, omitted variables bias was addressed in the study design | Healthcare outcomes and expenses | Higher COC was associated with lower risk of hospitalization (OR=0.37, CI:0.36-0.38) and ED visits (OR: 0.44, CI:0.43-0.45), and lower total healthcare expenses (β=-0.40, p<0.001) |
Gill 2000 ; USA [39] | CS | To examine the association of COC with ED visits | Delaware Medicaid database | 11,474 subjects | 1993-1994 | MMCI | 9/10 | Single year | Partly, reverse causality acknowledged as limitation | ED visits | High provider continuity is associated with lower single ED visit (OR=0.82, CI:0.70-0.95) and multiple ED visits (OR=0.65, CI:0.56-0.76) |
Hong 2013 ; Korea [54] | RC | To examine the effects of continuity of ambulatory care on health outcomes | Korean National Health Insurance Program | 68,469 patients; type 2 DM | 4 years | COCI | 9/9 | Unclear | Yes, addressed in the design and discussed in limitations | Hospitalization and healthcare costs | Lower COCI (compared to perfect COC of 1) was associated with increased risk of all-cause hospitalization (OR: 1.37, CI: 1.28-1.47) and healthcare costs (β= 0.037, P < 0.001) |
Hong 2010; Korea [55] | RC | To assess association of COC with health outcomes | Korea National Health Insurance Claims Database | 268,220 DM patients; 858,927 hypertension;129,550 asthma; 131,512 COPD (age: 65 to 84) | 2002-2006 | COCI | 8/9 | First 3 years | Yes, addressed in the design and discussed in limitations | Hospitalization, ED visits, healthcare costs | Low COC was associated with increased risk of hospitalization for patients with DM/ hypertension/ asthma/COPD (OR=1.47/ 1.31/2.07/1.99, CI:1.41-1.52/1.28-1.35/1.92-2.23/1.86-2.13, respectively) or ED visits (OR=1.41/ 1.45/2.25/1.77, CI:1.27-1.56/1.34-1.57/1.87-2.70/1.45-2.17, respectively) and increased healthcare costs (β= 0.130/0.116/0.025/0.123, respectively, P < 0.001 for all) |
Huang 2016; Taiwan [56] | RC | To examine whether continuity of ambulatory care can lower asthma-specific ED utilization | Taiwan National Health Insurance Dataset | 29,277 children; asthma patients (age 0 – 17 years) | 2006-2009 | COCI | 9/9 | Unclear | Yes, addressed in the design and in limitations | ED utilization | Low continuity of ambulatory care was associated with increased asthma-specific ED utilization (OR=1.38, CI:1.21-1.58) |
Hussey 2014; USA [18] | RC | To measure the association of COC with costs, rates of hospitalizations, emergency department visits, and complications for Medicare beneficiaries with chronic disease | US Medicare claims files for a 5% random sample of | 241,722 Medicare beneficiaries (CHF: 53,488; COPD: 76,520; DM: 166,654) [age > 65 years] | 2008-2009 | COCI | 8/9 | Unclear | Yes, discussed in limitations | Hospitalizations, ED visits, complications, and costs of care | Higher levels of COC among patients with CHF/COPD/DM was associated with lower rates of hospitalization (OR=0.94/0.95/0.95, CI:0.93-0.95/0.94-0.96/0.95-0.96, respectively) and ED visits (OR=0.92/0.93/0.94, CI:0.91-0.92/0.92-0.93/0.93-0.94, respectively), and lower episode costs by 4.7%/6.3%/5.1%, CI:4.4%-5.0%/6.0%-6.5%/5.0%-5.2%, respectively |
Jung 2018; Korea [57] | RC | To determine the association between COC and health outcome | Data from the Korea Health Insurance Review and Evaluation Service (HIRA) | 311,949 outpatients with knee osteoarthritis | 2014 | MFPC; MMCI; COCI | 7/9 | 9 months before hospitalizations | Yes, discussed in limitations | Hospitalization and medical expenses | Low COC was associated with increased risk of hospitalization (RR=27.17, CI:3.09-3.51) and medical expenses (β=0.677, P<0.001) |
Kao 2017; Taiwan [58] | RC | To investigate the relationship between COC and the risk of avoidable hospitalizations | Taiwan’s National Health Insurance claim data | 3,356 Elderly asthma patients | 2 years/ 2004-2013 | COCI | 9/9 | Single year | Partly, reverse causality is accounted in the design | Avoidable hospitalization | Low continuity was associated with increased risk of avoidable hospitalization (HR=2.68, CI:1.55-4.63) |
Kao 2017; Taiwan [58] | RC | To investigate whether high continuity of ambulatory asthma care reduces asthma-related ED visits | Taiwan Health Insurance claims Database 2010 | 3,395 Elderly asthma patients | 2 years/ 2004-2013/ | COCI | 9/9 | Single year | Partly, reverse causality is accounted in the design and limitations | Asthma-related ED visit | Low COC was associated with increased risk of ED visits (HR=2.11, CI:1.37-3.25) |
Kao 2019; Taiwan [27] | RC | To investigate associations between COC and ED visits and hospitalization for COPD or asthma | Taiwan National Health Insurance research database | 1141 asthma-COPD overlap patients aged ≥65 years | 2 years/2004-2013 | COCI | 8/9 | Single year | Partly, reverse causality is accounted in the design | Hospital admission for COPD or asthma, ED visits | The risk of hospital admissions for COPD or asthma for patients in the low and medium COC group was significantly higher than for those in the high COC group (aHR, 1.80; CI: 1.03–3.13; aHR, 1.72; CI: 1.04–2.83, respectively), which is true for ED visits: low COC group (aHR, 2.80; CI: 1.45–5.38); medium COC group (aHR, 2.69; CI: 1.47–4.93). |
Kim 2016; Korea [59] | RC | To examine the association of COC with complications health outcomes | Korean National Health Insurance Service (NHIS). | 715,053 hypertensive patients | 2007-2011 | MFPC; MMCI; COCI | 9/9 | 2 years before index event (outcomes one year later) | Partly, reverse causality accounted in the design | Hospitalization, ED visits, complications | Lower COC was associated with increased hospitalization (OR=1.25, CI:1.14-1.36), ED visits (OR=1.38, CI:1.13-1.70) and complications (OR=1.08, CI:1.02-1.13) |
Knight 2009; Canada [60] | RC | To investigate the relationship between continuity of family physician care and inpatient hospitalizations | Data from the Medical Care Plan (MCP) physician claims database and the Clinical Database Management System (CDMS) | 1,143 elderly DM patients (age ≥ 65 years) | 3 years /1998-1999 | COCI; UPC; SECON | 7/9 | Over 3 years after the index event | No | Inpatient hospitalizations | Higher continuity of family physician care was associated with reduced hospitalizations using UPC (OR=0.82, CI:0.68-0.98), COCI (OR=0.82, CI:0.69-0.97), and SECON (OR=0.75, CI:0.61-0.91) |
Lai 2016 ; Taiwan [61] | RC | To evaluate the COC and PIM effects | Taiwan Health insurance database | 823 DM patients with heart failure (age≥ 65 years) | 2005-2010 | COCI | 8/9 | Each year of follow-up | Yes, discussed in limitations | Hospital admissions and ED visits | Higher COC was associated with lower hospital admissions (OR=0.07, CI:0.05-0.10) and ED visits (OR=0.10, CI:0.07-0.13) |
Lei 2020a ; USA [42] | RC | To estimate the causal impact of COC on hospitalizations and different reasons for hospitalization | aggregate Veterans Health Administration and Medicare data at the veteran level using Veterans Affairs’ | 105,528 community-dwelling older veterans with dementia | 2014-2015 | COCI | 9/9 | Single year | Yes, accounted in the design and in limitations | Probability of hospitalization | 0.1 higher COC resulted in 2.4% (CI: 0.5%–4.4%) lower probability of hospitalization for all causes. COC was not associated with hospitalizations for ACSC. 0.1 higher COC resulted in 3.8% (CI: 2.1%–5.4%) lower probability of hospitalization for neuropsychiatric diseases/disorders. |
Lei 2020b ; USA [43] | RC | To estimate the causal impact of COC on total, institutional, and noninstitutional cost among | aggregate Veterans Health Administration and Medicare data at the veteran level using Veterans Affairs’ | 102 073 community-dwelling older veterans with dementia | 2014-2015 | COCI | 9/9 | Single year | Yes, accounted in the design and in limitations | Total costs and costs categories (acute, long-term care costs, ED costs) | 0.1 higher BBC resulted in $4045 (CI, $2171-$5919) lower total cost. 0.1 higher COC resulted in $1597 (CI, $688-$2506) lower acute inpatient cost, $119 ( CI, $64-$174) lower ED cost, $4368 (CI, $643-$8093) lower long-stay nursing home cost, $402 (CI, $113-$691) higher medical LTC cost, and $764 (CI, $460-$1067) higher social LTC cost |
Li 2019 ; Taiwan [62] | RC | To investigate whether COC is associated with healthcare outcomes and medical care use | Taiwanese National Health Insurance database | 4,007 patients with newly diagnosed diabetes | 2010-2012 | COCI, UPC | 7/9 | Over 3 years | Yes, discussed in limitations | number of hospital admissions, length of hospital stays, and number of ED visits | The high COCI and UPC groups had significantly lower probabilities of adverse outcomes: probability of hospital admissions (adjusted OR=0.623, CI: 0.543-0.716), probability of ED visits (adjusted OR=0.650, CI:0.570-0.741). high COCI group had a significantly lower incidence rate ratio (IRR) for the number of hospitalizations (IRR=0.75, CI: 0.67-0.83), length of hospitalizations (IRR=0.61, CI:0.52-0.72), and number of ED visits(IRR=0.68, CI: 0.62-0.75) |
Lin 2017; Taiwan [63] | RC | To examine the effects of high COC on the risk of avoidable hospitalizations | Longitudinal Health Insurance Database from Taiwan National Health Research Institute | 2,199 COPD patients | 2005 | COCI | 8/9 | Short term (1 year); long term (2 years) | Yes, accounted in the design | COPD-related avoidable hospitalization | Short-term COC: medium and low COC were associated with increased risk of avoidable hospitalizations, although significant for medium group only (OR=1.89, CI: 1.07-3.33). Long-term COC: medium and low COC were associated with increased risk of avoidable hospitalizations (OR=1.98, CI:1.0-3.94 and OR=2.03, CI:1.05-3.94, respectively) |
Lin 2015; Taiwan [64] | RC | To examine the association of COC with the risk of future hospitalization | Longitudinal Health Insurance Database from Taiwan National Health Research Institute | 3,015 COPD patients | 2005 | COCI | 9/9 | Over 2 years | Yes, discussed in limitations | COPD-related avoidable hospital admission | Lower COC was associated with a higher likelihood of COPD-related avoidable hospitalization (OR=2.29, CI:1.26-4.15) |
Lin 2010; Taiwan [65] | RC | To examine the association of discontinuity of care with the risk of hospitalization | Data from Taiwan National Health Research Institute (NHRI). | 6,476 DM patients | 1997-2002 | UPC | 9/9 | Over follow-up period | Yes, discussed in limitations | Diabetes-related- admission | Lower COC was associated with higher risk of short-term (OR=1.124, CI: 0.547-02.310) and long-term (OR=1.336, CI:1.019-1.728) ACSC admissions. |
Madison Hyer 2020; USA [66] | RC | To characterize the impact of continuity of care on perioperative outcomes, as well as on cost of care, among Medicare beneficiaries undergoing hepatopancreatic resection | Medicare claims data. | 25,698 Medicare beneficiaries who underwent a hepatopan-creatic surgical procedure | 2013-2017 | COCI | 7/9 | Single year | Partly, reversed causality accounted in design | Total costs of surgery, incidence of complications, length of stay, 30- and 90-day readmission and mortality | Among patients undergoing hepatic resection, an increase in COC of 0.2 was associated with decreased costs of 5.1% (CI: -6.3% to -3.8%) compared with a decrease of 2.5% (CI: -3.7% to -1.2%) among pancreatic resection patients. Higher COC was associated with lower 30-day readmission rate(hepatic:COC1st quartile: 13.5% versus COC4th quartile: 11.7%; pancreatic COC1st quartile: 18.6% versus COC4th quartile 16.3%) and lower 90-day readmission rate (hepatic: COC1st quartile: 20.8% versus COC4th quartile: 18.4%; pancreatic COC1st quartile: 27.1% versus COC4th quartile 24.1%)(both P <.05). |
Mainous III 1998; USA [67] | RC | To examine the association of (site and clinician) COC with the risk of future hospitalization | Delaware Medicaid Patients | 13,495 patients; ≥ 3 visits to ambulatory care (age: 0- 65 years) | 1993-1995 | UPC, Site Index, Clinician index | 7/9 | Single year | Partly, reversed causality accounted in design | hospitalizations | High COC with a clinician had lower odds of hospitalization (OR=0.75, CI: 0.66-0.87) than high site/low clinician COC. High site/low clinician COC group was not significantly different from low site/low clinician COC (OR=0.93, CI: 0.80-1.08) |
Menec 2005; Canada [40] | CS | To examine the association of COC with preventive healthcare and ED use | Administrative data; physician claims data and the Manitoba Immunization Monitoring System (MIMS) database. The Population Registry; Canada Census | 536,893 subjects; ≥1 physician contact | 1998-1999 | Proportion of total visits to family physicians (FPs) made to the same FP - majority-of-care rule | 7/10 | Over follow-up period | Yes, discussed in limitations | ED visits | Higher COC (>75%) was associated with reduced ED use among children (OR=0.85, CI: 0.78–0.92) and adults (OR=0.85, CI: 0.80–0.90) |
Menec 2006; Canada [68] | RC | To examine the relationship between COC and hospitalizations | Survey data linked to administrative physician billing data | 1863 subjects (age ≥ 67 years) | 1990-1991; 1996-1997 | Proportion of total visits to family physicians (FPs) made to the same FP - majority-of-care rule | 8/9 | Over 2 two-year periods (of follow-up) | No | Hospitalizations | High COC was associated with reduced odds of being hospitalized for ACSC (OR=0.67, CI:0.51-0.90) but not with hospitalizations for all conditions (OR=0.83, CI:0.67-1.01) |
Nam 2016; Korea [69] | RC | To examine the association between time-dependent COC and recurrent hospital admissions | Korean National Health Insurance Claims Database (KNHI) | 34,607 participants; hypertensive patients | 2011–2013 | COCI | 9/9 | over 2011-2012 (single year) | Partly, reversed causality accounted in design | Hospital admissions | Lower COC was associated with a higher risk of hospital admission (HR=1.42, CI: 1.10-1.83) |
Nyweide 2013; USA [70] | RC | To examine the relationship between COC and the risk of preventable hospitalization | Claims data FFS Medicare beneficiaries | 3,276,635 subjects (age > 65 years | 2007-2010 | COCI; UPC | 8/9 | cumulatively each succeeding month until the occurrence of the event | Yes, discussed in limitations | Preventable hospital admissions | Higher continuity of ambulatory care was associated with a lower rate of preventable hospitalization (HR=0.98, CI:0.98-0.99) |
Pennap 2020.; USA [71] | RC | To assess the patient-provider continuity of care (CoC) and compare the risk of psychiatric ED visits or hospitalization according to the CoC level. | Medicaid administrative claims data | 38,825 individuals, 3–16-year old with a first psychiatric diagnosis between 2009 and 2013 | 2007-2014 | Alpha Index | 8/9 | Over 2 years | Partly, reversed causality accounted in design | The risk of psychiatric ED visits or hospitalization | The odds of ED visits were higher among youths with low CoC (OR=1.27; CI: 1.13–1.41)or moderate CoC (OR=1.14;CI: 1.02–1.27) compared with those with high CoC Greater odds of psychiatric hospitalization related to low (OR= 1.17; CI: 1.06–1.29) or moderate CoC (OR=1.15; CI: 1.03–1.27) compared with high CoC. |
Pollack 2013; USA [41] | CS | To examine the association between patients’ care density and their healthcare costs | Data from 5 large commercial insurance plans | 9,596 patients (CHF); 52,668 patients (DM) [age ≥ 40 years] | 2009 | Care density | 7/10 | Between baseline and interview time (≥ 1 year apart) | No | Healthcare costs | Patients treated by sets of physicians who shared high numbers of patients (higher care density) tended to have lower costs (total: lower by 3’310$, p<0.001, inpatient: 2’563$, p=0.001) and rates of hospitalizations (83.4% of hospitalization in low density group, p<0.001). |
Pollack 2015; USA [72] | RC | To examine if care density is associated with measures of quality | Data from 3 large commercial insurance plans | 31,675 (CHF) ; 78,530 (COPD); 240,378 (DM) [age ≥ 40 years] | 2008-2009 | Care density | 7/9 | Over follow-up period | Partly, reversed causality accounted in design | 30-day readmission, quality indicators | Higher care density was associated with reduced rates of 30-day readmissions (OR=0.68, CI:0.48-0.97) |
Reddy 2018; USA [73] | RC | To examine the association of team-based care and COC on high-cost healthcare utilization | Medicare claims data for Veterans Affairs (VA) | 1,160,365 patients | 2012-2013 | UPC | 8/9 | Single year | Yes, discussed in limitations | ACSC hospitalizations, ED | Increasing COC by 10 percentage points was associated with lower hospitalizations (-4.5, CI [-5.3;-3.7]) and ACSC hospitalizations (-3.2, CI [-3.4;-2.9], but not significantly with ED visits (2.6, CI [-0.2; 5.4]). |
Romaire 2014; USA [20] | CS | To examine the effect of COC among beneficiaries who primarily see a PCP and those who primarily see a specialist | Medicare FFS claims data | 613,471 Medicare beneficiaries | 2007-2009 | Predominant Provider; UPC; COCI | 10/10 | Single year | Yes, discussed in limitations | All-cause hospitalizations, ACSC, all-cause ED visits, and ACSC ED visits, expenditures paid by Medicare | Regardless of specialty type of the predominant provider, higher continuity was associated with lower rates of all cause hospitalization (PCP: IRR=0.91, CI:0.90-0.93, Specialist: IRR=0.91, CI: 0.88-0.95), all-cause ED (PCP: IRR=0.85, CI:1.05-1.11, Specialist: IRR=0.85, CI: 0.83-0.88) and ACSC ED use (PCP: IRR=0.90, CI:0.89-0.92, Specialist: IRR=0.85, CI: 0.82-0.88), and lower expenditures for these services and total costs (PCP: β=-0.158, Specialist: β=-0.131, p<0.0001). |
Swanson 2018; Germany+Norway [74] | RC | To compare a social health insurance country (Germany) and a national health service country with gatekeeping and patient lists (Norway) on continuity of primary care for patients in terms of GP visits before and after their first hospitalization and compared the effect of COC on 30-day and one-year hospital readmission rates following hospital discharge | German insurance claims data and linked Norwegian national register data; the database for control and payment of reimbursements to health service providers (KUHR), the cause of death register (DÅR) and the GP database (Fastlegedatabasen) | 6,373 (Germany); 13,507 (Norway); COPD patients | 2009-2014 | COCI; UPC; SECON | 8/9 | Over 2 years before index event and 1 year after | Partly, reverse causality accounted in the design and limitations | Hospital readmissions within 30 and 365 days | Higher GP continuity in primary care was associated with reductions in hospital readmissions, regardless of systems’ set up and COC indices used. Germany 365 day readmission: COCI: OR=0.958, CI: 0.945-0.971, UPC: 0.951, CI: 0.934-0.969, SECON: OR=0.976, CI: 0.962-0.989. Norway 65 day readmission: COCI: OR=0.855, CI: 0.847-0.863, UPC: 0.808, CI: 0.799-0.817, SECON: OR=0.865, CI: 0.857-0.873. |
Vogt 2016 ; Germany [75] | RC | To investigate the relationship between provider continuity in ambulatory care and admissions | Scientific Research Institute of the regional health insurance funds ‘Allgemeine Ortskrankenkassen’ (AOKs) | 382,118 heart failure patients (age ≥ 35 years) | 2009-2011 | COCI; UPC; SECON | 9/9 | Over 2009 -2010 before hospitalization | Partly, reversed causality accounted in design | Hospitalizations | Higher COC among GP, internists and cardiologists was associated with a reduction in the risk of hospitalizations (COCI: OR=0.860, CI: 0.800-0.926, UPC: OR=0.834, CI: 0.758-0.918, SECON: OR=0.752, CI: 0.692-0.818). Higher COC with GPs only was associated with lower hospitalization using SECON (OR=0.874, CI: 0.799-0.957) but not the other COC indices |
Wang 2020 ;Taiwan [26] | RC | To investigate the relationships between COC and chronic conditions, how it impacted the number of outpatient visits, and risk factors of highly frequent uses | National Health Insurance Research Database (NHIRD) in Taiwan | 33,294 patients who had at least one internal medicine outpatient visit | 2007-2009 | COCI | 8/9 | Single year | Partly, reversed causality accounted in design | Risk of high medical utilization (>51 visits) | Patients in the low and moderate COCI groups had a significantly higher risk of the use than did the patients in the high COCI group (OR = 2.38, CI: 2.12–2.68 and OR = 1.96, CI: 1.74–2.21, respectively). Those who had severe comorbidities (OR = 3.03, CI: 2.84–3.23) were more likely to use outpatient care highly frequently. |
Worrall 2011 ; Canada [76] | RC | To examine the relationship between continuity of family physician care and all-cause mortality and acute hospitalizations | Sample from Newfoundland and Labrador portion of the National Diabetes Surveillance System (NDSS) database linked to provincial FFS physician billing database | 305 older people with DM (age ≥ 65 years) | 1998-1999 | UPC | 7/9 | Over 3 years | No | Death and hospitalization rate | Higher COC group had significantly lower likelihood of hospitalizations than lower COC group (54.5% and 67.5%, respectively, p=0.027). |
Yang 2020; Taiwan [77] | RC | To investigate the association of COC index (COCI) with medical costs and inpatient days, and investigate the possible clinical characteristics affecting the outcome | Taiwan’s National Health Insurance Research Database | 3234 patients aged 0 to 18 years with cerebral palsy catastrophic illness | 2000-2013 | COCI | 8/9 | Single year | No | Medical costs and the number of inpatient days over 5 years | Five-year inpatient days for a child in the low COC group were longer than for a child in the high COC group (8 days more, p < 0.001). Five-year medical costs for a child in the low COC group were higher than for a child in the high COCI group (US$1656 more, p = 0.016). For inpatient costs: US$1660 more, p = 0.002. |
Characteristics of the COC measures
Measure | Description | Range | Concept * | Formula | Interpretation | Number of papers using this index |
---|---|---|---|---|---|---|
1.Bice & Boxerman index (COCI) | the degree to which patient visits are distributed among different physicians | 0 (all visits to different providers) to 1 (all visits to the same provider) | dispersiona densityb | \(\large \frac{{\sum }_{i=1}^{M}{n}_{i}^{2}-N}{N(N-1)}\) | N-total number of visits, ni number of visits with provider i, M-total number of providers | 35 |
2.Usual Provider of Care (UPC) | the proportion of visits with a usual provider | 0 (no visits to the usual provider) to 1 (all visits to the usual provider) | density | \(\large \frac{{{max}_{i=1,\dots M} n}_{i}}{N}\) | N- total number of visits, ni – number of visits with provider i, M-total number of providers | 14 |
3.Most Frequent Provider Continuity (MFPC) (similar to UPC)** | defines the primary provider as the one seen most frequently | 0 (no visits to the provider) to 1 (all visits to this providermost frequent provider) | density | \(\large \frac{max({n}_{1},{n}_{2}\dots , {n}_{M})}{N}\) | N - total number of visits, ni - number of visits with provider i (i=1,…M), M-total number of providers | 2 |
4.Modified, Modified Continuity Index (MMCI) | the extent to which a patient concentrated her/his visits with the same healthcare provider | 0 (all visits to different providers) to 1 (all visits are to the same provider) | dispersion | \(\large \frac{1-\frac{M}{N+0.1}}{1-\frac{1}{N+0.1}}\) | M -number of different providers and N- total number of visits. | 5 |
5.Sequential Continuity of Care index (SECON) | the proportion of sequential visits that were with the same provider, i.e. same provider being seen at both the previous and current visits | 0 (no sequential visits to the same provider) to 1 (all sequential visits to the same provider) | sequential c | \(\large \frac{{\sum }_{j=1}^{n-1}{\text{c}}_{\text{j}}}{n-1}\) | if the visit j and the subsequent visit (j + 1) are to the same provider then cj = 1, and cj = 0 if otherwise; n-1- sequential pairs of visits | 5 |
6.Herfindahl Index | indicator of provider concentration, physician’s share of a patient’s visits. | 0 (less-dominant providers, not concentrated) to 1 (all visits to the same provider, very concentrated) | concentrationd | \(\large {\sum }_{i=1}^{M}{\left(\frac{{n}_{i}}{N}\right)}^{2}\) | ni - visits for patients to an individual provider i, and N - total number of visits, M-total number of providers | 1 |
7.Care density | reflect how frequently patient’s doctors collaborate/ share patients with one another | 0 (no collaboration/ no shared patients) to ∞ (extreme collaboration/ all patients of all doctors are shared) | density; level of shared patients between providers | \(\large \frac{{\sum }_{i=1}^{m}{\text{w}}_{\text{p,i}}}{{\text{M}}_{\text{p}}({\text{M}}_{\text{p}}-1)/2}\) | Mp - number of distinct doctors that patient p saw, m - total number of possible pairs of doctors, and wp,i - number of shared patients for each pair of doctors. The numerator is the total number of instances of patient sharing over a time period (e.g. a year) among a patient’s doctors. The denominator is the total number of pairs of doctors for that patient. | 3 |
8.Clinician Index | the proportion of visits with primary clinician out of all ambulatory visits | 0 (no visits to the primary clinician) to 1 (all visits to the primary clinician) | density | np/Na1 | np - number of ambulatory visits to a primary clinician and Na1 - number of ambulatory visits in the 1st year | 1 |
9.Site Index | the proportion of visits with primary site out of all visits | 0 (no visits to the primary site) to 1 (all visits to the primary site) | density | np/N1 | np - number of visits to a primary site and N1 - total number of visits in the 1st year | 1 |
Compound Indices | ||||||
10.Integrated Continuity of Care index (ICOC) | ICOC is a linear combination of three indices UPC, COC, SECON via Principal Component Analysis (PCA) | 0 (no visits to the same provider) to 1 (all visits to the same provider) | dispersion; density; sequence | ICOC=(β1 UPC+ β2COC+β3SECON)/ (β1+ β2+ β3) | ß1, ß2, and ß3 is the first principal component Eigenvector of the PCA result, which used the weighted means of the variables for each type of index | 1 |
11.Composite index (COMP) | Composite index, derived by adding the score values of the four commonly used COC indices (COCI, UPC; MMCI, SECON) and dividing by four | 0 (no visits to the same provider) to 1 (all visits to the same provider) | dispersion, density, sequence | (UPC+COCI+MMCI+SECON)/4 | UPC - usual provider of care index, COCI - Bice& Boxerman index, MMCI - Modified modified continuity index, SECON - sequential continuity index | 1 |
12.CI Alpha Index | weighted average of the concentration of providers seen and sequential continuity; KL represents Kullback-Leibler information index showing the degree of concentration relative to no concentration at all | 0 (a different provider is seen for each patient visit, maximum dispersion in both provider concentration and visit sequence) to 1(the same provider is seen at every visit) | concentration and sequence | CIα = αKL* + (1 - α) SECON, α ϵ [0, 1]; \(KL= log M+ {\sum }_{i=1}^{M}\frac{{n}_{i}}{M} \times log \frac{{n}_{i}}{M}\); and \(KL*=\frac{KL}{\mathit{log}M}\) | α is a predetermined weight that is applied to both KL* and SECON; M - total number of providers and ni is number of visits with provider i. | 1 |
Outcomes associated with COC
Outcomes | Total number of studies | Number of studies observing significant improvement in outcomes associated with improved COC |
---|---|---|
Costs | ||
All-cause and disease-specific hospitalizations | ||
Avoidable hospitalizations and ACSC | ||
ED use | ||
Hospital readmissions | ||
Outpatient visits use | 1 [26] | 1 [26] |