University:
Boston University
Course:
GMS BC 630 | Ethical & Policy Issues in Health and Medical Services Emergency Management
Academic year:

2023
Views:

478

Pages:

22
Author:

Tara Guerrero

WHAT PREDICTS MAJOR LEAGUE BASEBALL TEAM RELOCATIONS, 1950 – 1970? • • • • Race of residents living nearby Not attendance Not place in standings Not age of stadium Predicting hospital closings, 1990 - 2010 Predictor, 1990 values Signiﬁcance Beds (fewer) 0.000 Hospital fund balance / adj. daily census (lower) 0.000 Occupancy rate (lower) 0.004 Area percent black (higher) 0.008 Hospitals within 1 mile (more) 0.034 Case mix-‐adjusted cost / discharge (eﬃciency) 0.637 OperaQng margin 0.679 Model C-statistic = 0.819 Predicted Chance of Hospital Closing, 1990 – 2010 (Mean hospital – as function of mean 1990 characteristics) β Estimate Independent Variable (coefficient) Intercept -‐2.190 Beds 0.005 Area percent black -‐0.010 Occupancy rate 1.948 Hospitals i n 1 mile -‐0.168 Fund balance/adjusted census 0.004 Case mix-‐adjusted cost/discharge 0.000 Values for Prediction mean for mean hospital hospital 1.000 -‐2.190 328.7 1.772 28.7 -‐0.298 66.1 1.288 1.2 -‐0.202 $134,183 0.573 $6,462 0.090 Sum Exponential value of sum Predicted probability of survival 1.033 2.810 73.8% Predicted probability of closing mean hospital 26.2% Model C-statistic = 0.819 Predicted Chance of Hospital Closing, 1990 – 2010 (Mean hospital – as function of mean 1990 characteristics) β Estimate Independent Variable (coefficient) Intercept -‐2.190 Beds 0.005 Area percent black -‐0.010 Occupancy rate 1.948 Hospitals i n 1 mile -‐0.168 Fund balance/adjusted census 0.004 Case mix-‐adjusted cost/discharge 0.000 Values for Prediction mean for mean hospital hospital 1.000 -‐2.190 328.7 1.772 28.7 -‐0.298 66.1 1.288 1.2 -‐0.202 $134,183 0.573 $6,462 0.090 Sum Exponential value of sum Predicted probability of survival 1.033 2.810 73.8% Predicted probability of closing mean hospital 26.2% Model C-statistic = 0.819 Predicted Chance of Hospital Closing, 1990 – 2010 (At-risk hospital – predicted by riskiest quartile values of variables in 1990) Independent Variable Intercept Beds Area percent black Occupancy rate Hospitals i n 1 mile Fund balance/adjusted census Case mix-‐adjusted cost/discharge β Estimate (coefficient) -‐2.190 0.005 -‐0.010 1.948 -‐0.168 0.004 0.000 Riskiest Prediction for quartile 1990 hospital in values riskiest quartile 1.000 -‐2.190 176.0 0.949 45.4 -‐0.472 57.0 1.111 2.0 -‐0.336 $33,508 0.143 $4,769 0.067 Sum Exponential Predicted probability of survival -‐0.729 0.482 32.5% Predicted probability of closing hospital in riskiest quartile 67.5% Model C-statistic = 0.819 Predicted Chance of Closing between 1990 and 2010 Rises as Beds Fall and as Area Percent Black Rises Beds, 1990 Higher quartile Area Percent Black, 1990 Lower Mean quartile 600 433 329 5% 6.1% 11.8% 21.9% 38.9% 49.0% Mean 29% 7.7% 14.6% 26.4% 44.9% 55.1% Higher quartile 45% 9.0% 16.9% 29.9% 49.2% 59.4% 75% 11.8% 21.7% 36.7% 56.9% 66.5% 99% 14.7% 26.2% 42.7% 62.9% 71.8% Lower quartile 176 100 Of the 548 non-public hospitals with 50 or more beds that were open in 1990, 193 (35%) closed by 2010. Chance of closing was calculated from mean 1990 values of all significant variables—except Beds and Area Percent Black, which changed with the cell being calculated. Predicted Chance of Hospital Closing, 1980–2010 (Mean hospital – as function of mean 1980 characteristics) Independent Variable Intercept Beds Area percent black Medical school affiliation Occupancy rate Hospitals in 1 mile Fund balance/adjusted census β Estimate (coefficient) -3.441 0.004 -0.011 0.329 2.681 -0.190 0.010 Values for mean Prediction for hospital mean hospital 1.000 324.2 29.7 0.7 0.8 1.6 $56,470 Sum Exponential value of sum Predicted probability of survival Predicted probability of closing of mean hospital -3.441 1.452 -0.324 0.002 2.038 0.000 0.544 0.271 1.311 56.7% 43.3% Model C-statistic was 0.824. Of 608 non-public hospitals with 50 or more beds in 1980, 291 (47.9%) closed by 2010. Decade-by-decade Logistic Regression Results • All exclude public hospitals or < 50 beds • Backwards stepwise: 0=closed + 1=open • Panel 1 – six decades from 1936 to 2000 • Panel 2 – predicting from decade-old data • Panel 3 – multi-decade predictions Panel 1 – Decade-by-decade, 1936-50-60-70-80-90-2000 • All values for predictors are for the year at start of the period • A few consistent, standard-set predictors: – – – – Beds, Area percent black, teaching, occupancy rate • Fairly consistent results • C-statistics range from 0.760 to 0.836 Logistic Regression Summary by Decade, 1936 - 2000 Dependent Variable Alive 1936-‐50 3a 0=close/1=survive Independent Variables β Estimate Intercept 1.1327 BEDS37 0.0115 APB40 -‐0.0184 TCHNG37 1.2939 Model c-‐statistic 0.836 Dependent Variable Alive 1950-‐60 3b 0=close/1=survive Independent Variables β Estimate Intercept 1.8446 BEDS50 0.0063 APB50 -‐0.0243 Model c-‐statistic 27 close 461 survive P-‐value 0.0188 0.0159 0.0161 0.0174 Dependent Variable 3d Independent Variables Intercept BEDS70 APB70 OCCUP70 Model c-‐statistic 43 close 465 survive P-‐value <0.0001 0.0006 <0.0001 Dependent Variable 3e Independent Variables Intercept BEDS80 APB80 MSA80 β Estimate -‐2.531 0.00588 -‐0.01 4.1607 97 close 530 survive P-‐value 0.0007 <0.0001 0.0077 <0.0001 0.784 Alive 1980-‐90 β Estimate 0.0631 0.0072 -‐0.014 0.5126 111 close 507 survive P-‐value 0.8177 <0.0001 <0.0001 0.0166 0.776 Model c-‐statistic Dependent Variable Alive 1960-‐70 3c Independent Variables β Estimate Intercept -‐0.7793 BEDS60 0.00285 APB60 -‐0.0231 OCCUP60 4.1752 Model c-‐statistic Alive 1970-‐80 54 close 518 survive P-‐value 0.3449 0.0191 <0.0001 0.0004 Dependent Variable 3f Independent Variables Intercept BEDS90TF APB90 OCCUP90 0.817 Alive 1990-‐2000 111 close 446 survive β Estimate P-‐value -‐1.2439 0.0035 0.00756 <0.0001 -‐0.015 0.0001 1.9522 0.005 11 Panel 2 – Predicting from Decade-old Data • Values of independent variables are 10 years old at start of period – Closings 2000 – 2010 as function of 1990 values – Closings 1990 – 2000 as function of 1980 values • Compare standard-set with expanded-set of independent variables 2. Predicting from Decade-old Data Consistent standard-set predictors Expanded-set predictors Dependent Variable: Alive 2000-‐2010 82 close 3g =f(1990 values) 355 survive Independent Variables β Estimate P-‐value Intercept -‐0.5705 0.2243 BEDS90TF 0.00341 0.0001 OCCUP90 1.4668 0.0484 Dependent Variable: Alive 2000-‐2010 82 close 5 =f(1990 values) 355 survive Independent Variables β Estimate P-‐value Intercept -‐0.204 0.475 BEDS90 0.004 0.000 FBXAC90 0.004 0.000 Model c-‐statistic Model c-‐statistic 0.717 Dependent Variable Alive 1990-‐2000 107 close 3k =f(1980 values) 398 survive Independent Variables β Estimate P-‐value Intercept -‐2.4103 0.0009 BEDS80 0.00403 <0.0001 APB80 -‐0.0136 0.0009 MSA80 0.4057 0.0243 OCCUP80 3.5587 0.0005 Model c-‐statistic 0.792 0.746 Panel 3 – Multi-decade Predictions • Predicting 20 or 30 years of behavior • As function of values at beginning of period • Consistent standard-set predictors compared with expanded-set predictors 3a. Multi-decade predictions: 1990-2010 Consistent standard-set predictors Dependent Variable: 3i Independent Variables Intercept BEDS90TF APB90 OCCUP90 Model c-‐statistic Alive 1990-‐2010 β Estimate -‐2.0105 0.00541 -‐0.0117 2.2134 0.787 193 close 355 survive P-‐value <0.0001 <0.0001 0.0007 0.0004 Expanded-set predictors Dependent Variable: Alive 6 1990-‐2010 Independent Variables β Estimate Intercept -‐2.190 BEDS90 0.005 APB90 -‐0.010 OCCUP90 1.948 HM190 -‐0.168 FBXAC90 0.004 Model c-‐statistic 0.815 193 close 355 survive P-‐value 0.000 0.000 0.003 0.003 0.035 0.000 3b. Multi-decade predictions 1980 - 2000 Consistent standard-set predictors Dependent Variable: 3j Independent Variables Intercept BEDS80 APB80 MSA80 OCCUP80 Model c-‐statistic Alive 1980-‐2000 β Estimate -‐2.9883 0.00525 -‐0.0154 0.4964 3.1194 0.827 220 close 398 survive P-‐value <0.0001 <0.0001 <0.0001 0.001 0.0002 Expanded-set predictors Dependent Variable: Alive 2 1980-‐2000 Independent Variablesβ Estimate Intercept -‐3.5476 BEDS80 0.0053 APB80 -‐0.0153 MSA80 0.3740 OCCUP80 3.7380 HMI180 -‐0.2228 FBXAC80 0.0097 Model c-‐statistic 0.846 220 close 398 survive P-‐value 0.0000 0.0000 0.0000 0.0293 0.0000 0.0007 0.0001 3c. Multi-decade predictions, 1980 - 2010 Consistent standard-set predictors Dependent Variable: 3l Independent Variables Intercept BEDS80 APB80 MSA80 OCCUP80 Model c-‐statistic Alive 1980-‐2010 β Estimate -‐2.7837 0.00428 -‐0.0114 0.4825 2.102 0.805 291 close 317 survive P-‐value <0.0001 <0.0001 0.0005 0.0003 0.0093 Expanded-set predictors Dependent Variable: Alive 3 1980-‐2010 Independent Variablesβ Estimate Intercept -‐3.441 BEDS80 0.004 APB80 -‐0.011 MSA80 0.329 OCCUP80 2.681 HMI180 -‐0.190 FBXAC80 0.010 Model c-‐statistic 0.824 291 close 317 survive P-‐value 0.000 0.000 0.001 0.029 0.002 0.002 0.000 Characteristics of survivors and closings, 1990-2000, 1990-2010, and 2000-2010 Predicting closings, 1990 - 2000 Predictors 1990 mean values all 1990 still open close by hospitals 2000 2000 1990 values predict n = 612 Beds 90 n=499 p n=113 Predicting Closings, 2000 - 2010 Predicting closings 1990 - 2010 1990 mean values still open close by p 2010 2010 1990 mean values still open close by p 2010 2010 n = 404 n = 404 n = 198 n = 85 348.0 385.3 183.2 0.0000 413.4 216.3 0.0000 413.2 259.1 0.0000 2.1 2.0 2.3 0.0000 2.0 2.2 0.0000 2.0 2.2 0.0032 Area % Hispanic 90 15.7 15.8 15.0 0.6812 16.3 14.4 0.2910 16.2 14.8 0.5662 Area % minority 90 44.2 42.9 50.2 0.0170 42.7 47.2 0.0775 42.8 44.1 0.7173 Area % black 90 29.4 27.9 36.1 0.0077 27.3 33.6 0.0130 27.4 30.1 0.4392 Med sch affil 90 0.79 0.91 0.27 0.0000 0.98 0.40 0.0000 0.98 0.55 0.0001 COTH 90 0.32 0.37 0.09 0.0000 0.40 0.13 0.0000 0.40 0.19 0.0002 Age assets 90 8.13 8.16 8.01 0.7504 8.07 8.30 0.5547 8.07 8.63 0.2829 Cost, cmi-adjust 90 $4,890 $4,943 $4,656 0.3782 $4,975 $4,684 0.2858 $4,954 $4,728 0.5703 Cost, unadjusted 90 $6,644 $6,818 $5,882 0.0473 $6,973 $5,951 0.0096 $6,987 $6,091 0.1165 1.23 1.25 1.12 0.3816 1.26 1.08 0.1657 1.2 1.0 0.1883 67.3% 69.7% 56.7% 0.0000 71.4% 58.9% 0.0000 71.3% 61.8% 0.0000 Fund bal/adj cens 90 $133,280 $149,463 $62,393 0.0000 $162,753 $72,558 0.0000 $164,504 $107,236 0.0127 Total margin 90 Ownership 90 Hosps w/in 1 mile 90 Occupancy rate 90 Operating margin 90 0.68 1.57 -3.19 0.0000 2.10 -2.26 0.0000 2.04 -0.99 0.0052 -1.27 -0.66 -3.94 0.0074 -0.03 -3.88 0.0002 -0.14 -3.72 0.0097 Closed Brooklyn Hospitals St. Mary’s Interboro Brooklyn Jewish Shibe Park, Philadelphia Presbyterian Mt. Sinai Northeastern Hospital of Women’s Medical College Braves Field, Boston The wrong team left town. “Survival of the fattest.”