Acute Physical Health Shocks and Mental Health Care

Wei Song

Supervisors: Panos Kasteridis, Rowena Jacobs

Health Econometrics and Data Group

2nd October, 2024

Logo 1 Logo 2

[1.0] Background and motivation

  • Physical and mental health are closely related with biological and behavioural pathways.

    30% of people with LTCs have mental health problems;
    46% of people with mental health problems have LTCs.
    (Naylor et al., 2012)

  • More research focuses on how mental health impacts physical health. Reciprocal pathways and endogeneity bring about challenges in modelling impact.

[1.1] Acute phyiscal health shock

  • Acute myocardio infarction, cerebral infarction, and new diagnosis of cancer

  • Unanticipated timing helps with the endogeneity problem

  • The experience of an acute physical health shock are shown to negatively impact risk for depression and general anxiety disorder.

  • Would acute physical health shock impact mental health service utilisation? If so, would it be increased or decreased?

[1.2] Research question

  • How does the experience of acute physical health shocks impact the utilisation of specialised mental health care?

  • ... for individuals who are existing specialised mental health care users

    • Possible challenges

  • How to capture acute physical health shock?
  • How to deal with different timing for treatment and outcome measures?
  • Would variations in shock intensity complicate measurement, as individuals are displaced due to hospitalisation for shock related physical care?

[2.0] Data and methods

  • Hospital Episode Statistics (HES): Admitted Patient Care (APC)
  • Mental Health Services Data Set (MHSDS)
Data Chart
  • Population: users of specialised mental health care in 2017/18
  • Treatment: first experience of heart attack, shock, or first diagnosis of cancer
  • Outcome: days spent in mental health care

[2.1] Mental health care pathways in England

  • Primary care: GP referrals, IAPT (talking therapies)

  • Community care

  • Hospital care at NHS Mental Health Trust (secondary and tertiary)

  • Specialist mental health care delivered at physical health setting (integrated care)

[2.2] Data and methods (cont.)

Data Chart

[2.3] Data and methods (cont.)

Propensity score matching

Sex Age group Mental health cluster Ethnicity IMD quartile Elixhauser comorbidity index Pre-treatment utilisation of mental health care

[3.0] Econometric setup

Data Chart

\( y_{it} = \alpha_0 + \alpha X_i + \beta d_i + \gamma \lambda + \color{orange}{\delta d_i \lambda} \) + \( u_{it} \)

[3.1] Multi-timepoint design

Data Chart

Each \( \lambda \) represents a three-month period

[3.2] Stacked DiD

Data Chart

\( y_{it} = \alpha_0 + \alpha X_i + \beta d_i + \sum\limits_{t=2}^{T} \gamma_t \lambda_t + \sum\limits_{t=T_s}^{T} \delta_t(d_i \lambda_t) + u_{it} \)

[3.3] Treatment intensity

Data Chart

\( \small y_{it} = \alpha_0 + \alpha X_i + \sum\limits_{k=1}^{K} \beta_k d_{ik} + \sum\limits_{t=2}^{T} \gamma_t \lambda_t + \sum\limits_{k=1}^{K} \sum\limits_{t=T_s}^{T} \delta_{tk}(d_{ik} \lambda_t) + u_{it} \)

[4.0] Results for inpatient: covariate balance

Data Chart

[4.1] Inpatient (cont.)

Data Chart
  • Total number of inpatient bed-days: 7.4m for 2016/17, 7.5m for 2019/20
  • Average number of inpatient bed-days: 7 to 8 days per quarter
  • Higher utilisation in July to December time

[4.2] Inpatient (cont.)

Data Chart

[4.3] Inpatient (cont.)

Data Chart

[4.4] Inpatient (cont.)

[4.5] Inpatient (cont.)

Men

[5.0] Results for outpatient: covariate balance

Data Chart

[5.1] Outpatient (cont.)

Data Chart
  • Total number of days spent in community care: 151m 2016/17, 135m 2019/20
  • Average number of days range from 20 to 25 days per quarter

[5.2] Outpatient (cont.)

Data Chart

[5.3] Outpatient (cont.)

Data Chart

[5.4] Outpatient (cont.)

[5.5] Outpatient (cont.)

Men

[6.0] Results, mental health care in physical setting (HES IP)

Data Chart

[6.1] HES IP

Men

[6.2] HES OP

Men

[7.0] Modelling incremental cost

  • Chapter 1, survey data that links experience of PHS to increased risk for depression
  • Chapter 2, admin data that links experience of PHS to altered utilisation in community and inpatient mental health care
  • Chapter 3, a model that facilitates interpolation of cost difference

[7.1] State transition

[7.2] Data structure

Population

\[ %\mathbf{X} = \begin{bmatrix} %\mathbf{x}_1 \\ %\mathbf{x}_2 \\ %\vdots \\ %\mathbf{x}_N %\end{bmatrix} = %\left. %\begin{matrix}\, % \smash{ % \underbrace{ \begin{bmatrix} x_{1,1} & x_{1,2} & \cdots & x_{1,K} \\ x_{2,1} & x_{2,2} & \cdots & x_{2,K} \\ \vdots & \vdots & \ddots & \vdots \\ x_{N,1} & x_{N,2} & \cdots & x_{N,K} \end{bmatrix} % }_{\text{variables}} % } % \vphantom{ % \begin{matrix} % \smash[b]{\vphantom{\Big|}} % 0\\0\\0\\vdots\\0 % \smash[t]{\vphantom{\Big|}} % \end{matrix} % } %\,\end{matrix} %\right\rbrace{} \]
  • Age
  • Gender
  • Socioeconomic Status
  • Comorbidity
  • Care Cluster

Dynamic
State

\[ \begin{aligned} \mathbf{S} = \{ s_{i,t}^{(l)} \mid &i = 1,\dots,N; \\ &t = 1,\dots,T; \\ &l = 1,\dots,L \} \end{aligned} \]
  • Health shock status
  • Utilisation
  • QALY

Outcome
Aggregation

\[ %\mathbf{Y} = \begin{bmatrix} y_{1,1} & y_{1,2} & y_{1,3} \\ y_{2,1} & y_{2,2} & y_{2,3} \\ \vdots & \vdots & \vdots\\ y_{N,1} & y_{N,2} & y_{N,3} \end{bmatrix} \]
  • Total utilisation days
  • Total QALY
  • Care Cluster
  • Currencies: HRG & Cluster

[8.3] Model workflow

[8.0] Discussion

Displacement

Removal of utilisation during quarter of shock If present, could be picked up by TE heterogeneity of shock intensity

P-trend assumption

Defining core population

PHS and SMI

Stacked and staggered design?

Treatment intensity not fully reflected with current approach?

Reduction in utilisation: supply or demand driven?

Thank you.