Outline Comparative Analysis of Referral Costs in GP Practices Maria C. Portela Emmnuel Thanassoulis Sofia Nogueira Silva Ano lectivo 2008/2009 Maria Conceição Silva Portela • • • • • • Objectives of the study Data and Variables Weight restrictions DEA models Results on Aggregate total referral cost model Results on Disaggregate total referral cost model • Conclusion Ano lectivo 2008/2009 Objectives • In the UK GP practices are responsible for the health assurance of the individuals that are registered within the practice. • Referral costs depend crucially on the decisions made by the GP practice, including the choice of inpatient or outpatient referral and the choice of hospital for the referral, as treatment costs are largely standardised through • Most studies have focused on analysing the efficiency of GP practices providing the health service (i.e. translate resources into outcomes (or intermediate outputs)); • Here we focus on the efficiency of GP practices in purchasing the hospital treatment for their patients Ano lectivo 2008/2009 Maria Conceição Silva Portela Maria Conceição Silva Portela Data and Variables Inputs Outputs Model 1 Total costs (inpatient referral costs) + outpatient List of population enrolled in the practice, divided into homogeneous groups Model 2 Cost of referrals for inpatient care List of population enrolled in the practice, divided into homogeneous groups Cost of referrals for outpatient care Volume of referrals Price of referrals - Data on 75 GP practices Ano lectivo 2008/2009 Maria Conceição Silva Portela Data and Variables Data and Variables • The population characteristics were measured through: – Number of people in certain age bands; – Index of multiple deprivation (IMD), where we considered the number of people in each age band that lived in an area with na IMD above a certain percentile (50%, 60%, 70%, 80%, and 90%) – Several specifications of DEA models and regression analysis were tried to arrive at the final specification of outputs Ano lectivo 2008/2009 Maria Conceição Silva Portela Band code Age/gender Band 10 age < 3 3.09% Band 20 age >= 3 and age < 18 17.58% Band 30 age >= 18 and age < 35 and sex = Male 10.23% Band 35 age >= 18 and age < 35 and sex = Female 9.84% Band 40 age >= 35 and age < 48 and sex = Male 10.48% Band 45 age >= 35 and age < 48 and sex = Female 9.92% Band 50 age >= 48 and age < 70 27.11% Band 60 age >= 70 11.74% Number of people in these age bands below and above the 80% Percentile were considered as the outputs (16 in total) Ano lectivo 2008/2009 Weight Restrictions - Costs for people above the percentile (more deprived) are higher; - Some age bands can be considered as having similar costs and other have clearly higher costs (looking at values below the percentile) Ano lectivo 2008/2009 Weighte d average inpatient cost per person Weighted Average inpatient cost ( below 80% percentile) Weighted Average inpatient cost (above 80% percentile) Band Age/gender Band 10 age < 3 173.57 174.81 157.12 Band 20 Age [3, 18[ 72.02 71.88 74.26 Band 30 age [ 18, 35[ Male 71.92 71.09 85.38 Band 35 age [ 18, 35[ Female 235.58 231.035 303.78 Band 40 age [ 35, 48[ Male 107.99 103.36 198.15 Band 45 age [ 35, 48[ Female 209.30 205.55 287.66 Band 50 Age [ 48, 70[ 278.29 276.16 331.95 Band 60 age >= 70 742.56 739.72 817.19 Maria Conceição Silva Portela Maria Conceição Silva Portela DEA Models Weight restictions were imposed to reflect that: Similar costs: Band 30, 20, 40; Bands 10, 45, 35 Average percentage in this band in the GP practices • Aggregate total referral cost model s m s Max ∑ u r y ro | ∑ u r y rj − ∑ vi xij ≥ 0, r =1 i =1 r =1 m ∑v x i io i =1 = 1, Au ≥ 0, u r , vi ≥ 0 • Disaggregate total referral cost model n n m Min c = ∑ pio xi | ∑ λ j xij ≤ xi , ∑ λ j y rj − A −1 w ≥ y ro , λ j , xi ≥ 0 j =1 j =1 i =1 Technical efficiency Ano lectivo 2008/2009 Allocative efficiency Maria Conceição Silva Portela DEA Models Results • Price efficiency models (Tone and Tsutsui, 2007) n Min ρ | ∑ λ j CijT ≤ ρ CioT , j =1 Price efficiency n ∑λ j =1 n Min ∑i Ci | ∑ λ j CijT ≤ Ci , j =1 Allocative efficiency ∑C ∑C i T ij i ij Ano lectivo 2008/2009 ∑ρ C ∑C * × i i T ij T ij × ∑C ∑ρ C * i y rj ≥ y ro , λ j ≥ 0 n ∑λ j j =1 * ij i j = T ij y rj ≥ y ro , λ j ≥ 0 ∑C ∑C i * ij i ij Maria Conceição Silva Portela Ano lectivo 2008/2009 Results – aggregated model average Number eff minimum • Cost CRS Efficiency Cost VRS Efficiency Scale Efficiency 78.29% 85.41% 91.69% 6 16 6 61.36% 63.00% 71.06% • Returns to scale Maria Conceição Silva Portela Results – Aggregate model • Benchmarking practices were identifed Returns to scale were constant or increasing up to a list size of 2808, and after 4687 people, only DRS were identified. List size Average CRS St dev CRS <3000 83.64% [3000,5000[ 76.23% [5000,7000[ [7000,10000[ >10000 Ano lectivo 2008/2009 Average VRS St dev VRS Average scale eff St dev scale 0.129 88.18% 0.113 79.78% 75.89% 0.103 79.15% 0.117 75.75% 0.074 0.125 95.14% 0.084 16 0.102 95.38% 0.032 17 83.63% 0.100 90.65% 0.033 10 86.93% 0.090 90.73% 0.053 16 88.20% 0.082 85.93% 0.036 16 Maria Conceição Silva Portela N Ano lectivo 2008/2009 Maria Conceição Silva Portela Results – disaggregate model • Average efficiency scores • Cost benchmark GP practices Cost eff CRS Tech eff CRS Alloc eff CRS Scale eff Cost eff average 79.00% 84.95% 92.95% 94.05% min 58.33% 62.65% 77.67% 69.36% Number Eff 5 10 5 10 Tech eff VRS Alloc eff VRS 86.48% 90.35% 95.61% 59.71% 63.11% 79.24% VRS 16 VI 21 16 • Scale efficiency results from aggregate model were confirmed; • Benchmark practices for cost and technical efficiency measures were identified Ano lectivo 2008/2009 Results – disaggregate model Maria Conceição Silva Portela G5 G13 G40 G44 Average G45 G9 402 1463 2437 359 1028.2 858 789 VO 2079 6968 11958 2245 5062.2 4275 2996 VI/VO 0.193 0.210 0.204 0.160 0.203 0.201 0.263 PI 1241.97 1307.53 1244.66 1203.02 1261.62 1011.15 1034.38 88.91 86.84 87.33 97.12 89.36 96.77 93.24 13.97 15.06 14.25 12.39 14.12 10.45 11.09 2780 9004 14696 2786 6411 3519 3484 PI/PO List Pop Ano lectivo 2008/2009 Results- Disaggregate model Maria Conceição Silva Portela Results – Disagregate model • Cases of volume increase or decrease to achieve cost efficiency (after technical efficient targets) % TVI/VI >1 # cases average <=1 # cases average =1 # cases average Ano lectivo 2008/2009 % TVO/VO 13 43 1.006 1.033 46 16 0.906 0.904 16 16 1.000 1.000 Maria Conceição Silva Portela Price efficiency results: average # Eff Ano lectivo 2008/2009 Tech Eff VRS Price Eff VRS Mix Eff VRS Cost Eff VRS 90.35% 97.32% 96.02% 84.54% 21 35 15 15 Maria Conceição Silva Portela Results – disaggregated model Results II Price efficient benchmarks of GP 43 Ano lectivo 2008/2009 Maria Conceição Silva Portela Conclusion • Some GP practices were found cost inefficient, i.e. facing excessive costs in relation to the other practices that serve similar populations: – (i) scale issues - smaller GP practices appear more efficient than larger practices but at the same time have more variability in cost efficiency, which may be a result of less stable case-mixes over time; – (ii) technical efficiency- most GP practices presented excessive volume of inpatient and outpatient referrals given their population characteristics; – (iii) price efficiency- some GP practices face much lower average prices and therefore achieve lower costs by means of lower prices; – (iv) mix efficiency- some GP practices may reduce their costs by adjusting the volume mix to closely match the relationship between their prices . Ano lectivo 2008/2009 Maria Conceição Silva Portela Ano lectivo 2008/2009 Maria Conceição Silva Portela