Choosing Wisely: Trend Graphs
Bob Johnson
April 22, 2016
Each of the figures below show a similar pattern. There is an increase in rates in early Sep-15 followed by a dip. A smaller dip occurs around Jan-16. The Mar-16 rates are close to the peak of the Sep-15 rates.
### BURN team
LAB="BURN"
Team = subset(FileMonth,Care.Team==LAB)
attach(Team)
### BMP
BMP.Rate = 100* BMP/Census
SLine=Smoother("BMP",LAB) 
MonData(BMP)## Month Count Census Rate
## 6 June 185 269 69
## 5 July 218 403 54
## 1 August 104 205 51
## 10 September 192 251 76
## 9 October 101 346 29
## 8 November 161 309 52
## 2 December 239 363 66
## 4 January 203 356 57
## 3 February 144 255 56
## 7 March 251 334 75### CBC
CBC.Rate = 100* CBC/Census
SLine=Smoother("CBC",LAB) 
MonData(CBC)## Month Count Census Rate
## 6 June 152 269 57
## 5 July 214 403 53
## 1 August 93 205 45
## 10 September 148 251 59
## 9 October 84 346 24
## 8 November 156 309 50
## 2 December 193 363 53
## 4 January 166 356 47
## 3 February 183 255 72
## 7 March 228 334 68### BMP & CBC combined
BMP.CBC.Rate = 100* BMP/Census
SLine=Smoother("BMP.CBC",LAB) 
MonData(BMP.CBC)## Month Count Census Rate
## 6 June 337 269 125
## 5 July 432 403 107
## 1 August 197 205 96
## 10 September 340 251 135
## 9 October 185 346 53
## 8 November 317 309 103
## 2 December 432 363 119
## 4 January 369 356 104
## 3 February 327 255 128
## 7 March 479 334 143In these figures we see a sharp rise around the beginning of Sep-15. For BMP, that rise came from a consistent lower rate. The peak slowly lowers to a mid-range rate by Mar-16. For CBC the trend is more volatile but also has a peak around mid-Sep-15 and returns to mid-range by Mar-16.
### Geriatrics team
LAB="GERIATRICS"
Team = subset(FileMonth,Care.Team==LAB)
attach(Team)
### BMP
BMP.Rate = 100* BMP/Census
SLine=Smoother("BMP",LAB) 
MonData(BMP)## Month Count Census Rate
## 6 June 316 510 62
## 5 July 276 511 54
## 1 August 355 581 61
## 10 September 368 491 75
## 9 October 405 625 65
## 8 November 418 586 71
## 2 December 405 612 66
## 4 January 418 635 66
## 3 February 363 580 63
## 7 March 396 652 61### CBC
CBC.Rate = 100* CBC/Census
SLine=Smoother("CBC",LAB) 
MonData(CBC)## Month Count Census Rate
## 6 June 226 510 44
## 5 July 283 511 55
## 1 August 233 581 40
## 10 September 284 491 58
## 9 October 264 625 42
## 8 November 283 586 48
## 2 December 279 612 46
## 4 January 220 635 35
## 3 February 245 580 42
## 7 March 292 652 45### BMP & CBC combined
BMP.CBC.Rate = 100* BMP/Census
SLine=Smoother("BMP.CBC",LAB) 
MonData(BMP.CBC)## Month Count Census Rate
## 6 June 542 510 106
## 5 July 559 511 109
## 1 August 588 581 101
## 10 September 652 491 133
## 9 October 669 625 107
## 8 November 701 586 120
## 2 December 684 612 112
## 4 January 638 635 100
## 3 February 608 580 105
## 7 March 688 652 106Here we are seeing a rise around Sep-15, but there are two other rises and dips over the observed time window. For CBC, the higher rise occurs around early Dec-15.
### MICU NP team
LAB = "MICU NP"
Team = subset(FileMonth,Care.Team==LAB)
attach(Team)
### BMP
BMP.Rate = 100* BMP/Census
SLine=Smoother("BMP",LAB) 
MonData(BMP)## Month Count Census Rate
## 6 June 997 749 133
## 5 July 940 747 126
## 1 August 880 739 119
## 10 September 1061 806 132
## 9 October 228 209 109
## 8 November 262 222 118
## 2 December 227 222 102
## 4 January 251 218 115
## 3 February 289 213 136
## 7 March 293 229 128### CBC
CBC.Rate = 100* CBC/Census
SLine=Smoother("CBC",LAB) 
MonData(CBC)## Month Count Census Rate
## 6 June 520 749 69
## 5 July 529 747 71
## 1 August 526 739 71
## 10 September 633 806 79
## 9 October 144 209 69
## 8 November 189 222 85
## 2 December 156 222 70
## 4 January 141 218 65
## 3 February 117 213 55
## 7 March 154 229 67### BMP & CBC combined
BMP.CBC.Rate = 100* BMP/Census
SLine=Smoother("BMP.CBC",LAB) 
MonData(BMP.CBC)## Month Count Census Rate
## 6 June 1517 749 203
## 5 July 1469 747 197
## 1 August 1406 739 190
## 10 September 1694 806 210
## 9 October 372 209 178
## 8 November 451 222 203
## 2 December 383 222 173
## 4 January 392 218 180
## 3 February 406 213 191
## 7 March 447 229 195I have not attempted to model seasonal treads or autocorrelated trends. This would require a longer window of time, though autocorrelation could be modeled.