The analysis of the data on the understaffing of COVID ICUs, mortality and Greek hospital occupancy

Data sources

For the analysis regarding the (under)staffing of the COVID-19 ICUs, have been used primary data from 51 COVID-19 ICUs in 41 hospitals across Greece, provided by the Hellenic Society of Intensive Care Medicine (HSICM) to the iMEdD Lab. Their recording period extends from January 28, 2021 to February 18, 2021. The primary data are included in a .pdf file and presented in a five-column table, as described below.

• Hospital: each hospital under study
• COVID ICU beds: ICU under study and the corresponding number of beds
• Doctors: number of doctors on duty per ICU under study –in many cases special notes are included on the number of both auxiliary and seconded doctors, and those serving in the COVID-19 ICUs due to the current needs, who are not specialized in intensive care (residents, those transferred from other departments and so on).
• Nurses: number of nurses on duty per ICU under study –as with the medical personnel, in many cases, special notes are included on the number of nurses recorded in the total amount of personnel, who offer auxiliary services although they are not specialized in intensive care (transferred from other departments, trainees, etc.)
• Health District: the health district to which the respective hospital under study belongs

These data, provided by the HSICM to the iMEdD Lab, have subsequently been supplemented with additional information for the Department of Respiratory Deficiency (DRD), the 1^st ICU and the 2^nd ICU (former COVID) of Papanikolaou Hospital in Thessaloniki, according to the statement of the Department Manager of the Intensive Care Unit and Coordinating Director of the 2^nd ICU, Nikos Kapravelos, at the iMEdD Lab.

The following information is used for data analysis, regarding the relationship between local mortality and occupancy in the Greek Health System per Health District, during the second “wave” of the pandemic in Greece:

• The latest data on the available and occupied beds, in ICUs and in total, per Health District, which were disclosed by the Ministry of Health, as an informal briefing via e-mail to the Press, on November 20, 2020.
• The data on the geographical distribution of total losses, since the beginning of the pandemic, published by the iMEdD Lab in mid-December: they regard the distribution of 3,214 deaths per Regional Unit, a sample that corresponds to 87% of the 3,687 total losses by December 14, 2020.
• Calculated population figures per Regional Unit on January 1^st, 2019, which the Hellenic Statistical Authority (ELSTAT) provided, upon request, to the iMEdD Lab.

Processing and analysis of data

Using the Tabula tool, we converted the table with the primary data for the beds and the respective medical personnel per ICU under study, which we had originally received from the HSICM in a .pdf file to a .csv file. Then, with the use of spreadsheets, we proceeded to produce two different files for the medical and nursing personnel respectively. Each of them, after cleaning the quantitative data and removing the textual information, was developed as shown in the tables below.

The above table, which was the main body of data on which the analyses on the staffing of COVID ICUs with nursing personnel are based, includes the following variables:

• ΥΠΕ: reference health district
• ype_label: label for the reference Health District in Greek
• ype_label_en: label for the reference Health District in English
• hoverlabel: label for the reference ICU in Greek –in the form of “Hospital, ICU”, for the cases of multiple ICUs under study in the same hospital
• hoverlabel_en: label for the reference ICU in English –also in the form of “Hospital, ICU”, for cases of multiple ICUs under study in the same hospital
• icu_beds: number of beds in the reference COVID ICU
• nurs_thresh: estimated required total number of nurses per ICU under study, based on the minimum requirement for operation which is specified as four –trained in intensive care and emergency medicine– nurses per ICU bed, according to the unanimous decision of the Central Board of Health (KESY, 2016)
• nurses_total: total amount of nursing personnel on duty, as stated in the primary data provided to us
• nurses_net: estimated amount of nursing personnel specialized in intensive care –is obtained by subtracting the personnel that is supposed to serve in support, without being specialized in intensive care (transferred from other departments, trainees, etc.) from the total number of nurses on duty, provided that this is clearly stated in the special notes of the units in the data provided to us
• nurses_total_beds_ratio: ratio of total nursing personnel per bed
• nurses_net_beds_ratio: ratio of estimated specialized nursing personnel per bed

The above table, which was the main body of data on which the analyses on the staffing of COVID ICUs with medical personnel are based, in addition to the aforementioned regarding Health Districts, ICUs under study and corresponding available beds, includes the following variables:

• doct_thresh: estimated required total number of doctors per ICU under study. This is calculated with the following formula doct_thresh = 6+((icu_beds-8)/3) and is based on the unanimous decision of KESY (2016): according to the latter, at least six doctors, with the required specialization for intensive care, for eight ICU beds are “necessary for the safe and unobstructed operation of the unit”, while for ICUs larger than eight beds, the proportion of the number of doctors should be one doctor for every three extra beds. “In ICUs with four to six beds, the medical personnel cannot be comprised by less than six people, in order to cover on call duties” the same decision states.
• doct_total: total amount of medical personnel on duty, as stated in the primary data provided to us
• doct_net: estimated amount of permanent medical personnel –is obtained by subtracting the personnel that is supposed to serve in support, without being specialized in intensive care (auxiliary doctors, seconded employees, transferred from other departments, trainees, etc.) from the total number of doctors on duty, provided that this is clearly stated in the special notes of the units in the data provided to us
• doct_total_beds_ratio: ratio of total medical personnel per bed
• doct_net_beds_ratio: ration of estimated permanent medical personnel per bed

The analysis of the sample in terms of staffing of the units was, therefore, guided by the decision of KESY (2016) and is presented with a focus, on the one hand, on the individual ratios of healthcare personnel per bed at the ICUs under study and, on the other hand, on the consequent median ratio of healthcare workforce per bed at Health District level. Absolute numbers should be read with extreme caution in an ever-changing situation, the data of which are not sufficiently communicated by the Authorities. In particular, the competent bodies have never, to date, disclosed data on the staffing of hospitals during the pandemic, adding to the fragmentary, periodic, update regarding the existing infrastructure: the Ministry of Health proceeded to regular briefings regarding the available and occupied beds, per Health District, for COVID-19 cases only in the period from November 13 to November 20, 2020. There has been no regular release of data since, and there has been no response to the repeated requests from the iMEdD Lab to the Ministry of Health.

Part of the overall data processing and analysis was done using Python and the open library pandas.

Visualization of the analysis

The diagrams for the visualization of the analysis were created using the open source library Plotly Python.

Particularly for the visualization of the staffing of COVID-19 ICU beds with medical personnel, it was deemed appropriate to be done in the form of a range plot, to make obvious whether each of the ICUs included in our sample meets the minimum requirement for operation or is falling short, according to the information provided. For this purpose, firstly were created scatter plots, depicting the minimum required number of doctors and the declared number of doctors on duty per ICU under study. Subsequently, an equal number of lines were created, which have as starting and ending points the said numbers of doctors, minimum required and on duty respectively. Thus, the distance between the two, indicates the degree of meeting the needs or understaffing.