11  Health profile (1)

11.1 Putting it all together

Health profiles combine multiple indicators to allow comparison, usually between population or administrative units e.g. regions, hospitals, ethnic groups, age groups and provide tools to facilitate a range of comparison.

11.1.1 Workflow

11.1.1.1 Preparing data

To generate profile-style presentations per-indicator datasets need to be configured in a standard way and pooled. The minimum meta-data (variable names) for each indicator would be:

Meta-data for profile data
Indicator name
Area name
Age group
Gender
Value
Count (numerator)
Denominator
Metric type e.g. proportion, rate
Unit e.g. per 100,000, %
Lower uncertainty interval
Upper uncertainty interval
Uncertainty e.g. 95%, 99.8% 
flowchart TD
A[Configure per-indicator data] --> B[bind datasets]
Figure 11.1: Proposed workflow for creating health profiles

UK health profiles (for example, Public Health Outcome Framework) offer a wide range of comparison:

  • Between and within areas

  • Between age groups and genders

11.2 Visualisation

Comparison Visualisation Comment
Single area, multiple indicators Spine or bullet chart Indicator values are scaled so that they can be directly compared. Single area (unit) values are usually encoded as a dot or line, plotted on a bar representing the range of values for all areas / units.
Multiple areas, single indicator Column / bar chart Plots can be ordered and plotted horizontally to enable rapid appraisal of relative performance. Confidence intervals can be added
Indicator pairs / multiple indicators Scatter plot or scatter plot matrices May include smooths (lines representing fit to the data - common options are linear, loess and generalised additive models)
Spatial variation Choropleth maps Requires boundary files and GIS tools
Overview Heatmap / tile chart Colour cells with statistical significance categories (e.g. high / low)

We will take the toolkit for plotting provided by the UKHSA which generates the charts used in the online version if the PHOF as a basis for plotting profiles. This is available at https://github.com/ukhsa-collaboration/fingertipscharts.

Note: in the PHOF statistical significance is defined by whether the global average value (e.g. KSA) is contained with the confidence limits of the unit values (e.g. ’Asir region). It does not calculate limits for the global value. The rationale for this is explained in Public Health Data Science (2018)

Code 
## install fingertipscharts
## 

if(!require("fingertipscharts")) devtools::install_github("https://github.com/ukhsa-collaboration/fingertipscharts", force = TRUE)

library(fingertipscharts)

To use fingertipscharts data has to be in the right format.

Bar charts can be created using the compare_areas function

Code 
smok_rates |>
    compare_areas(area = region, value = value, lowerci = lowercl, upperci = uppercl, title = "Rate of attendance at smoking clinics 18-44: F")

flu_coverage |>
    filter(gender == "female" & age == "0-18") |>
    compare_areas(area = region, value = value, lowerci = lowercl, upperci = uppercl, fill = gender, title = "Flu vaccination coverage 0-18; F")
(a) Smoking rates
(b) Paediatric flu coverage
Figure 11.2: Using `compare-areas

11.3 Data overview

PHOF uses a heatmap (known as a “tartan rug” to display area by indicator summaries. The term tartan rug is used to refer to the colour scheme (red-amber-green) to denote statistical significance. The

Code 
prof_data <- bind_rows(smok_rates, flu_coverage) |>
    select(-c(11:13)) 

## calculate ksa values
ksa_values <- prof_data |>
    group_by(indicator) |>
    reframe(ksa_values = 100000 * sum(n, na.rm = TRUE) / sum(pop))

## join to data
prof_data <- prof_data |>
    left_join(ksa_values)

## code for statistical significance

prof_data <- prof_data |>
    mutate(significance = case_when(lowercl > ksa_values ~"high" ,
                                     uppercl < ksa_values~ "low",
                                    TRUE ~ "ns"))

head(prof_data) |>
    gt::gt()
Table 11.1: Required data format example
indicator region age gender n pop value lowercl uppercl timeperiod ksa_values significance
smoking 18-44 'Asir 18-44 female NA 366324 NA NA NA 2023 7.618658 ns
smoking 18-44 Al Bahah 18-44 female 28 61969 45.183882 30.017729 65.30573 2023 7.618658 high
smoking 18-44 Al Hudud ash Shamaliyah 18-44 female 4 70149 5.702148 1.553643 14.59976 2023 7.618658 ns
smoking 18-44 Al Jawf 18-44 female 23 103536 22.214495 14.077587 33.33412 2023 7.618658 high
smoking 18-44 Al Madinah al Munawwarah 18-44 female NA 387944 NA NA NA 2023 7.618658 ns
smoking 18-44 Al Mintaqah ash Sharqiyah 18-44 female NA 877403 NA NA NA 2023 7.618658 ns

11.3.1 Heatmap

Code 
prof_data |>
    ggplot() +
    geom_tile(aes(region, indicator, fill = significance)) +
    scale_fill_manual(values = c("red", "darkgreen", "goldenrod")) +
    scale_x_discrete(position = "top") +
    theme(axis.text.x = element_text(angle = 90, hjust = 0)) +
    labs(y = "", x = "")

11.3.2