Challenges to assessing professional identity in medical students: a tale of two measures

Figures & data

Figure 1. The Professional Identity as a Physician (PPI) measure (A) and the Identity Integration (IdIn) measure (B).

Table 1. Hypotheses for validity evidence.

Hypotheses applying to both PPI and IdIn.

H1: PPI and IdIn scores will have a strong positive association. Rationale: The constructs represented by PPI and IdIn scores were expected to be similar but not identical.

H2. PPI and IdIn scores will increase across years of training. Rationale: As identity formation is a developmental process^10, we expected increases in scores for both measures with increases in training and/or maturation. (Increasing median scores would hold promise for being able to measure change in PIF with the scores.)

H3. Scores on a survey item addressing a related construct (indicate agreement with the statement: ‘What I do personally reflects on me as a professional’) will be positively related to PPI and IdIn scores. Rationale: Students who have a stronger identity with physicians and/or students with greater integration of personal/professional identities would have stronger agreement with this item; the associations should be stronger with increased level of training.

H4. Scores on a differently-formatted survey item addressing a related construct (yes/no to ‘A medical student should be held to higher standards than other types of students regarding the image he/she portrays on his/her personal [social media] page.’) would be positively related to PPI and IdIn scores. Rationale: If the pictorial response option format of the PPI and IdIn items did allow measurement of professional identity, higher scores on both PPI and IdIn would be associated with a ‘yes’ response (different response format) to this item.

Hypothesis only for Physician Professional Identity measure (PPI).

H5. There will be no relationship between PPI scores and responses to an unrelated but identically-formatted item about another identity construct. Rationale: It is reasonable to expect no statistical relationship between scores on identification with a religious group and PPI scores collected with the pictorial response format if that shared response format has no effect on the scores. A strong relationship between items directed at unrelated identity constructs, would suggest that the pictorial response format might be exerting response demand, calling into question the score validity.

Hypothesis only for Identity Integration measure (IdIn).

H6.There will be a negative relationship between IdIn scores and level of agreement with a related but logically-reversed survey item (‘My professional identity is different than my personal identity’). Rationale: If a negative association of moderate size is observed, concerns about the impact of response format on IdIn scores are reduced.

Table 2. Response distribution for 3473 students at 8 schools.

MSY	SchoolA	SchoolB	SchoolC	SchoolD	SchoolE	SchoolF	SchoolG	SchoolH	Total N
1	101/206	214/228	178/212	80/178	182/230	48/150	178/234	167/257	1148/1695
	49%	94%	84%	45%	79%	32%	76%	65%	68%
2	109/195	145/158	178/225	112/175	156/229	33/150	173/240	152/234	1058/1606
	56%	92%	79%	64%	68%	22%	72%	65%	66%
3	69/160	48/133	NC	NC	170/230	26/153	160/235	196/236	669/1148
	43%	36%			74%	17%	68%	83%	58%
4	51/150	39/98	NC	116/176	NC	30/143	170/230	192/237	598/1033
	34%	40%		66%		21%	74%	81%	58%
Total	330/711	446/616	356/437	308/529	508/690	137/596	681/940	707/964	3473/5482
	46%	72%	81%	58%	74%	23%	72%	73%	63%

Legend: Topline in each cell contains: number of responses/total number surveyed; lower line is the cohort’s response rate. (NC = Not Collected)

Figure 2. For the PPI item, the response pattern for the five options on the scale in Figure 1 varies across cohorts of medical students.

Figure 3. For the IdIn item, the response pattern for the five options on the scale in Figure 2 varies across cohorts of medical students.

Table 3. Association (Spearman’s rho statistics) of PPI and IdIn scores with responses to ‘What I do personally reflects on me as a professional’ (Hypothesis 3).

Measure	MS1	MS2	MS3	MS4
PPI	.01	.02	.03	.04
IdIn	.34	.31	.33	.34

All correlations p < .05.

Table 4. Association (Cramer’s V statistics) of PPI and IdIn scores with yes/no response to the statement that medical students should be held to higher standards than other students for social-media behavior (Hypothesis 4).

Measure	MS1	MS2	MS3	MS4
PPI	.09	.11*	.15*	.13*
IdIn	.25**	.17**	.21**	.27**

*p < .05

**p < .0001

Table 5. Summary of validity evidence for PPI and IdIn scores.

Hypothesis	Evidence for PPI scores	Evidence for IdIn scores	Summary for validity hypotheses
H1	None	None	Unconfirmed
H2	Modest	Unclear	Modest for PPI
H3	None	Moderate	Moderate for IdIn
H4	Modest	Moderate	Some evidence for both measures
H5	Moderate	– –	Confirmed
H6	– –	Strong	Confirmed