Human Capital Depreciation and Efficiency in Surgical Care (Text Version)

On September 21, 2011, Jason Hockenberry made this presentation at the 2011 Annual Conference. Select to access the PowerPoint® presentation (150 KB). Plugin Software Help.

Slide 1

Human Capital Depreciation and Efficiency in Surgical Care

Jason Hockenberry, PhD*, and Lorens Helmchen, PhD

* The authors have benefitted from collaboration and conversations with Peter Cram, MD, MBA, and Saket Girotra, MD.

Slide 2

Acknowledgements

This research was supported by grant number 1 R03 HS019743-01 (Principal Investigator [PI]: Hockenberry) from the Agency for Healthcare Research and Quality. The content is solely the responsibility of the presenter and does not necessarily represent the official views of the Agency for Healthcare Research and Quality.

The data used in this presentation are from the Pennsylvania Health Care Cost Containment Council (PHC4). This analysis was not prepared by PHC4. It was performed by the authors listed above. PHC4, its agents and staff bear no responsibility or liability for the results of the analysis, which are solely the opinion of the authors.

The authors have no conflicts of interest to declare.

Slide 3

Background

Organizational forgetting and human capital effects have received more attention from health economists recently.

• Gaynor, Seider and Vogt (2005).
• Gowrisankaran, Ho and Town (2006).
• Huckman and Pisano (2006).
• Hockenberry, Lien and Chou (2008).
• Sfekas (2009).
• David and Brachet (2009).

Slide 4

The Theory

Human Capital is accumulated through experience and education, interacts with endowments and is aggregated within organizations:

• Human Capital accumulation leads to increases in productivity.
• Breaks in production can lead to the depreciation of this capital (so-called forgetting) and reductions in productivity.

Slide 5

Model of Surgical Outcomes

Image: Two line graphs compare labor hours versus units produced under the Traditional Learning Curve and P(mortality) versus number of surgeries performed under the "Learning" Curve in Surgery.

What do surgeons produce? Surgeries, or something else?

Slide 6

Model of Surgical Outcomes

Consider the following:

1. Outcome= f(X,MDQuality,e).

Slide 7

Model of surgical outcomes

Consider the following:

1. Outcome= f(X,MDQuality,e).
2. Mortality_ijht = b₀ + b₁ X_iht + b₂ Phys. Quality_jt + e_ijht.

Slide 8

Model of Surgical Outcomes

Consider the following:

1. Outcome= f(X,MDQuality,e).
2. Mortality_ijht = b₀ + b₁ X_iht + b₂ Phys. Quality_jt + e_ijht.
3. Phys. Quality_jt = a₀ + a₁Physician Vol_jt-1 + a₂ t-(t-1) + ρ_j + u_ijht.

Slide 9

Model of Surgical Outcomes

Consider the following:

1. Outcome= f(X,MDQuality,e).
2. Mortality_ijht = b₀ + b₁ X_iht + b₂ Phys. Quality_jt + e_ijht.
3. Phys. Quality_jt = a0 + a1Physician Vol_jt-1 + a₂ t-(t-1) + ρ_j + u_ijht.

t-(t-1) = Temporal distance between procedures.
ρ_j = Physician fixed effect which captures the endowment.

Slide 10

Model of Surgical Outcomes

Consider the following:

1. Outcome= f(X,MDQuality,e).
2. Mortality_ijht = b₀ + b₁ X_iht + b₂ Phys. Quality_jt + e_ijht.
3. Phys. Quality_jt = a0 + a1Physician Vol_jt-1+ a₂ t-(t-1) + ρ_j + u_ijht.

So by substitution we get:

4. Mortality_ijht = d₀ + d₁ X_iht + d₂MDVol_jt-1+ d₃ t-(t-1)+ ρ_j + v_ijht.

Slide 11

Model of Surgical Outcomes

Consider the following

1. Outcome= f(X,MDQuality,e).
2. Mortality_ijht = b₀ + b₁ X_iht + b₂ Phys. Quality_jt + e_ijht.
3. Phys. Quality_jt = a0 + a1Physician Vol_jt-1+ a₂ t-(t-1)+ ρ_j + u_ijht.

By substitution we get:

4. Mortality_ijht = d₀ + d₁ X_iht + d₂MDVol_jt-1+ d₃ t-(t-1)+ ρ_j + v_ijht.

And there are, of course, always arguments about whether we have the salient parts of this included in these models.

Slide 12

Procedure of Interest

Examining a procedure used to treat Coronary Artery Disease (CAD)

• Percutaneous Coronary Interventions (PCI):
  • Actual procedure usually involves a single physician.
  • Often performed in emergent situations with little time for planning.

Slide 13

Data

• Source: Pennsylvania Health Care Cost Containment Council (PHC4).
• All inpatient admission claims within PA for the years 2006Q3-2010Q2.
• These data were augmented with variables calculating time (in number of days) between procedures where the physician was listed as the operator.

Slide 14

General Estimation Strategy

m_ijkt = β₀ + β₁D_jt + β₂V_jkt + β₃S_jt + β₄H_kt + β₅X_it + ν_ijkt

• m = mortality (≤1 day, in-hospital).
• D = measure of temporal distance of last surgery of surgeon j.
• V = vector containing volume of both surgeon j and hospital k.
• S = physician j's characteristics.
• H = hospital k's characteristics.
• X = patient i's characteristics.

Slide 15

Temporal Distance Measures

• A continuous covariate for temporal distance is not very informative.
• We define temporal distance indicators:
  • 0-2 days (ref).
  • 3-7 days.
  • 8-14 days.
  • 15+ days.
• We examine both the days since any OR and the days since the specific procedure.

Slide 16

Select Provider Characteristics

Slide 17

Select provider characteristics

Slide 18

Outcomes of PCI patients

Standard errors clustered at the physician level in parentheses
* p < 0.10, ** p < 0.05, *** p < 0.01.

Slide 19

Outcomes of PCI Patients Treated by High Volume Physicians

Standard errors clustered at the physician level in parentheses
* p < 0.10, ** p < 0.05, *** p < 0.01.

Slide 20

Does Surgeon Human Capital Depreciate?

• Survival after surgery appears to be negatively associated with temporal distance to an extent.
• The question is the 'root' of this effect.
  • Cognitive processes?
  • Manual dexterity?
  • Team coordination/mindfulness?

Slide 21

Resource Use

• We are also thinking about what temporal distance does to resource use.
• Increased temporal distance could increase resource use because of labor-capital tradeoffs.
• On the other hand it could reduce resource because more anomalies go unnoticed and therefore untreated, reducing resource use.

Slide 22

Resource use

P-values derived from standard errors clustered at the physician level in parentheses
* p < 0.10, ** p < 0.05, *** p < 0.01.

Slide 23

Simulation-Back of the Envelope

About 94 lives would have been preserved over 4 years in PA if all PCI patients were treated by those with higher levels of human capital (i.e., those operating w/ a 0-2 day temporal distance).

Extending these lives would have led to $117.5 M in total charges in treating PCI patients (about a 1.22% increase), or a cost of life extended of about $1.24M.

Slide 24

Limitations and Extensions

• We have access to dates but not time of day of procedures.
• We do not know the reason for these breaks from the OR.
• Further work is needed to ascertain the nature of this effect.