Table 3.
Unadjusted ORs and 95% CIs for the Associations Between Pathologist Practice of Assurance Behaviors and Influence Toward a More Severe MSL Diagnosis Due to Concerns About Medical Malpractice (n = 207)a
| How Do Medical Malpractice Concerns Affect Your Own Practice When Interpreting MSLs? | All Pathologists, No. (%) | Influence Toward a More Severe Diagnosis, No. (%) | No Influence on Diagnosis, No. (%) | OR (95% CI)b | P Value |
|---|---|---|---|---|---|
| Total | 207 (100.0) | 57 (27.5) | 150 (72.5) | ||
| 1. I request additional slides cut from the block | .017 | ||||
| No | 27 (13.0) | 1 (3.7) | 26 (96.3) | Reference | |
| Yes | 180 (87.0) | 56 (31.1) | 124 (68.9) | 8.02 (1.46-44.11) | |
| 2. I recommend additional surgical sampling | .047 | ||||
| No | 70 (33.8) | 13 (18.6) | 57 (81.4) | Reference | |
| Yes | 137 (66.2) | 44 (32.1) | 93 (67.9) | 2.03 (1.01-4.07) | |
| 3. I request second opinions | .10 | ||||
| No | 16 (7.7) | 1 (6.3) | 15 (93.8) | Reference | |
| Yes | 191 (92.3) | 56 (29.3) | 135 (70.7) | 4.31 (0.74-24.92) | |
| 4. I order additional tests such as IHC and or molecular tests | .16 | ||||
| No | 78 (37.7) | 17 (21.8) | 61 (78.2) | Reference | |
| Yes | 129 (62.3) | 40 (31.0) | 89 (69.0) | 1.59 (0.83-3.05) | |
| Composite variable for assurance behaviors | |||||
| Practiced at least one of four assurance behaviors (strongly agree vs all other responses) | .002 | ||||
| No (disagree/slightly disagree, agree/slightly agree) | 131 (63.3) | 26 (19.8) | 105 (80.2) | Reference | |
| Yes (strongly agree) | 76 (36.7) | 31 (40.8) | 45 (59.2) | 2.76 (1.47-5.15) | |
CI, confidence interval; IHC, immunohistochemistry; MSL, melanocytic skin lesion; OR, odds ratio.
aPercentages may not total 100 due to rounding. Except where noted, “no” is defined by Likert scale responses disagree (slightly disagree, disagree, strongly disagree) and “yes” is defined by agree (slightly agree, agree, strongly agree).
bFirth’s bias reduction method based on penalized likelihood (addresses issues of separability, small sample sizes, and bias of the parameter estimates). Wald χ2 statistics were used to test the effect each independent coefficient had on the direction of influence.