Abstract:
This dissertation consists of three essays on the economics of health-care delivery in hospitals. The rst two essays estimate the impact of doctor-patient demographic concordance on a doctor's decision making for diagnostic resources and medical treat- ments. Demographic concordance occurs when a doctor and patient have the same ethnic group and/or gender. The third essay estimates a relationship between ward- level nursing hours and a patient's health outcome. These three papers use de- tailed data obtained by me from a hospital. Ethics approval to use this data was granted from the New Zealand Health and Disability Council (Reference number: NTX/11/EXP/029). The rst essay estimates a relationship between gender and/or ethnic concordance between a doctor and patient and the amount of diagnostic tests ordered during a hospital stay. Diagnostic test orders have increased in many developed countries. For example, in the United States the cost of `unnecessary' diagnostic tests and pro- cedures has recently been estimated at between USD 200 to 250 billion per annum (Berwick and Hackbarth, 2012; Thompson, 2011). Therefore, ways to reduce unnec- essary diagnostic test ordering is of interest to health policy makers. I test whether doctors order higher or lower amounts of diagnostic tests when they treat patients with the same demographic features, relative to when they treat patients with no shared demographic features. I nd a statistically signi cant reduction in laboratory and radiology tests when a doctor treats a patient of the same gender and/or ethnicity relative to demographically discordant patients. Assuming demographic concordance variables are exogenous, I suggest two reasons for a reduction in diagnostic test orders. The rst reason is an information gain in demographically concordant consultations. Because information on a patient's health status comes from doctor-patient consul- tation and diagnostic tests, a reduction in diagnostic test orders suggests a doctor has obtained better quality information from consultation. An improved consultation could include gains in communication, and/or the physical exam. The information gain hypothesis only holds if preferences for the amount of laboratory and radiology tests do not change in demographically concordant relative to discordant pairs. The second reason for a reduction in test orders is if demographically concordant patients and/or doctors prefer to order fewer diagnostic tests, for example by choosing a less aggressive treatment plan. Unlike the United States, I do not expect litigation and insurance arrangements to explain a reduction in diagnostic tests, because health care is publicly provided and doctors are not at a personal risk of litigation in the hospital. The second essay estimates a relationship between doctor-patient ethnic concor- dance and a women's likelihood of having an emergency caesarean procedure. Studies have documented variation in caesarean procedure rates across ethnicity groupings within a country (Rumball-Smith, 2009; von Katterfeld et al., 2011; Vangen et al., 2000; Getahun et al., 2009). This paper makes a novel contribution to literature explaining ethnic-based variation in caesarean rates by investigating the e ect of provider-patient ethnic concordance or discordance on the decision to have an emer- gency caesarean. An emergency caesarean is decided after a women has gone into labour, and women who receive a planned caesarean are excluded from my sample. Di erences in the unobserved health status of women in ethnically concordant and discordant groups is therefore not expected to explain my results, because all women in the patient sample have been considered physically able to undergo a natural birth. I use the three largest casemanager ethnicity groupings in my data; European, Indian and Asian. I nd that Asian women with an Asian casemanager are on average 6% (p = :0001) less likely to have an emergency caesarean compared to an Asian women treated by a European or Indian casemanager1. Ethnic concordance for European and Indian patients is statistically insigni cant. I suggest three explanations for why Asian women are less likely to receive an emergency caesarean when treated by an 1A casemanager could be a midwife, sta nurse or doctor that is primarily responsible for a patient's care in hospital. Asian casemanager. These are a reduction in maternal distress, clinical uncertainty and/or patient preference for an emergency caesarean. Two primary reasons for ethnic-based variation in caesarean rates are di erences across ethnicity groupings in unobserved patient health characteristics and preferences for caesarean procedures. My result suggests that ethnicity-speci c health characteristics and preferences do not fully explain a higher caesarean rate for Asian women in New Zealand. The third essay estimates a relationship between ward-level hours of nursing sta and a patient's health outcome. Patient health outcomes are mortality and length of ward stay. There is a large body of empirical literature on the relationship between hospital nurse-to-patient ratios and patient health outcomes (for reviews see; Lang et al. (2004); Kane et al. (2007)). This paper contributes to this empirical literature in three ways: by using a detailed nursing sta dataset, using a novel instrumental variable for nursing hours and by considering the separate e ect of nursing and patient hours in a ward on a patient's health outcome. My instrumental variable is the amount of sick and bereavement leave taken by nurses on a ward. Initially, there is a statistically signi cant positive relationship between nursing hours on a ward and a patient's likelihood of mortality. After instrumentation, nursing hours on a ward has a negative, but statistically insigni cant, e ect on the likelihood of mortality. A patient's length of stay is modeled with a competing risk survival model. Discharge home is the main outcome. Competing risks are transfer to another health-care facility and in-hospital mortality. My main result is that cumulative exposure to higher patient hours on a ward is associated with a longer hospital stay in 16 out of 20 wards. An explanation for this is that increased demand by other patients on xed hospital resources, such as medical equipment and doctor and nurse time, lowers the ability to deliver timely hospital health care. As a result, patients stay in hospital longer to receive the health care they need. This information could be useful for hospital administrators, because it suggests improving patient ow through a hospital during high demand times could reduce the average length of stay.